Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+res0.png filter=lfs diff=lfs merge=lfs -text
+table_drawn_bbox_with_extra.png filter=lfs diff=lfs merge=lfs -text
+unitable/website/unitable-demo.gif filter=lfs diff=lfs merge=lfs -text
+unitable/website/unitable-demo.mp4 filter=lfs diff=lfs merge=lfs -text
+unitable/website/wandb_screenshot.png filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,18 @@
+
+unitable/experiments/unitable_weights/** 
+
+res/** 
+
+TestingFiles/**
+TestingFilesImages/**
+
+# python generated files
+__pycache__/
+*.py[oc]
+build/
+dist/
+wheels/
+*.egg-info
+
+# venv
+.venv

.gitlab-ci.yml

@@ -0,0 +1,33 @@
+variables:
+  GIT_STRATEGY: fetch
+  GIT_SSL_NO_VERIFY: "true"
+  GIT_LFS_SKIP_SMUDGE: 1
+  DOCKER_BUILDKIT: 1
+
+stages:
+  - build
+
+image_build:
+  stage: build
+  image: docker:stable
+  before_script:
+    - docker login -u gitlab-ci-token -p $CI_JOB_TOKEN http://$CI_REGISTRY
+  script: |
+    CI_COMMIT_SHA_7=$(echo $CI_COMMIT_SHA | cut -c1-7)
+    DATE=$(date +%Y-%m-%d)
+    docker build --tag $CI_REGISTRY_IMAGE/$CI_COMMIT_BRANCH:latest \
+                 --tag $CI_REGISTRY_IMAGE/$CI_COMMIT_BRANCH:$CI_COMMIT_SHA_7 \
+                 --tag $CI_REGISTRY_IMAGE/$CI_COMMIT_BRANCH:$DATE \
+                 -f Dockerfile .
+    docker push $CI_REGISTRY_IMAGE/$CI_COMMIT_BRANCH:latest
+    docker push $CI_REGISTRY_IMAGE/$CI_COMMIT_BRANCH:$CI_COMMIT_SHA_7
+    docker push $CI_REGISTRY_IMAGE/$CI_COMMIT_BRANCH:$DATE
+  # Run only when Dockerfile has changed
+  rules:
+    - if: $CI_PIPELINE_SOURCE == "push"
+      changes:
+        - Dockerfile
+      # Set to `on_success` to automatically rebuild
+      # Set to `manual` to trigger the build manually using Gitlab UI
+      when: on_success 
+      allow_failure: true

.vscode/settings.json

@@ -0,0 +1,2 @@
+{
+}

Dockerfile

@@ -0,0 +1,55 @@
+ARG BASE_IMAGE="nvidia/cuda:12.2.2-devel-ubuntu22.04"
+
+FROM ${BASE_IMAGE}
+ARG HOMEDIRECTORY="/myhome"
+ENV HOMEDIRECTORY=$HOMEDIRECTORY
+
+USER root
+RUN apt-get update && \
+   apt-get install -y  --no-install-recommends  \
+    curl \
+    python3 \
+    python3-pip \
+    python3-dev \
+    poppler-utils \
+    gcc \
+    git \
+    git-lfs \
+    htop \
+    libgl1 \
+    libglib2.0-0 \
+    ncdu \
+    openssh-client \
+    openssh-server \
+    psmisc \
+    rsync \
+    screen \
+    sudo \
+    tmux \
+    unzip \
+    vim \
+    wget && \
+    wget -q  https://github.com/justjanne/powerline-go/releases/download/v1.24/powerline-go-linux-"$(dpkg --print-architecture)" -O /usr/local/bin/powerline-shell && \
+    chmod a+x /usr/local/bin/powerline-shell
+
+RUN ln -s /usr/bin/python3 /usr/bin/python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt --extra-index-url https://download.pytorch.org/whl/cu117
+
+
+# setup ssh
+RUN ssh-keygen -A
+RUN sed -i 's/#*PermitRootLogin prohibit-password/PermitRootLogin yes/' /etc/ssh/sshd_config
+EXPOSE 22
+
+# Make the root user's home directory /myhome (the default for run.ai),
+# and allow to login with password 'root'.
+RUN echo 'root:root' | chpasswd
+RUN sed -i 's|:root:/root:|:root:/myhome:|' /etc/passwd
+
+ENTRYPOINT sudo service ssh start && /bin/bash
+
+
+
+
+

README.md

@@ -1,12 +1,118 @@
 ---
-title: Alps
-emoji: 👀
-colorFrom: purple
-colorTo: green
+title: alps
+app_file: app.py
 sdk: gradio
 sdk_version: 4.44.0
-app_file: app.py
-pinned: false
 ---
+# Alps
+
+Pipeline for OCRing PDFs and tables
+
+This repository contains different OCR methods using various libraries/models.
+
+## Running gradio:
+`python app.py` in terminal
+
+
+## Installation : 
+Build the docker image and run the contianer 
+
+Clone this repository and Install the required dependencies:
+```
+pip install -r requirements.txt --extra-index-url https://download.pytorch.org/whl/cu117
+
+apt install weasyprint
+
+```
+Note: You need a GPU to run this code.
+
+## Example Usage 
+
+Run python main.py inside the directory. Provide the path to the test file (the file must be placed inside the repository,and the file path should be relative to the repository (alps)). Next, provide the path to save intermediate outputs from the run (draw cell bounding boxes on the table, show table detection results in pdf), and specify which component to run.
+
+outputs are printed in terminal
+
+```
+usage: main.py [-h] [--test_file TEST_FILE] [--debug_folder DEBUG_FOLDER] [--englishFlag ENGLISHFLAG] [--denoise DENOISE] ocr
+
+```
+Description of the component:
+
+### ocr1 
+
+ocr1
+Input: Path to a PDF file
+Output: Dictionary of each page and list of line_annotations. List of LineAnnotations contains bboxes for each line and List of its children wordAnnotation. Each wordAnnotation contains bboxes and text inside.
+What it does: Runs Ragflow textline detector + OCR with DocTR
+
+Example: 
+```
+python main.py ocr1 --test_file TestingFiles/OCRTest1German.pdf --debug_folder ./res/ocrdebug1/ 
+python main.py ocr1 --test_file TestingFiles/OCRTest3English.pdf --debug_folder ./res/ocrdebug1/ --englishFlag True
+```
+
+### table1
+Input : file path to an image of a cropped table 
+Output: Parsed table in HTML form
+What it does: Uses Unitable + DocTR
+
+```
+python main.py table1 --test_file cropped_table.png --debug_folder ./res/table1/ 
+
+```
+
+### table2 
+Input: File path to an image of a cropped table
+Output: Parsed table in HTML form
+What it does: Uses Unitable
+
+```
+python main.py table2 --test_file cropped_table.png  --debug_folder ./res/table2/ 
+
+```
+### pdftable1
+Input: PDF file path
+Output: Parsed table in HTML form
+What it does: Uses Unitable + DocTR
+
+
+```
+python main.py pdftable1 --test_file TestingFiles/OCRTest5English.pdf  --debug_folder ./res/table_debug1/ 
+
+python main.py pdftable3 --test_file TestingFiles/TableOCRTestEnglish.pdf  --debug_folder ./res/poor_relief2
+```
+
+
+### pdftable2 : 
+Input: PDF file path
+Output: Parsed table in HTML form
+What it does: Detects table and parses them, Runs Full Unitable Table detection
+
+```
+python main.py pdftable2 --test_file TestingFiles/OCRTest5English.pdf --debug_folder ./res/table_debug2/ 
+```
+
+
+### pdftable3 
+Input: PDF file path
+Output: Parsed table in HTML form
+What it does: Detects table with YOLO, Unitable + DocTR
+
+
+
+### pdftable4 
+Input: PDF file path
+Output: Parsed table in HTML form
+What it does: Detects table with YOLO, Runs Full doctr Table detection
+
+python main.py pdftable4 --test_file TestingFiles/TableOCRTestEasier.pdf --debug_folder ./res/table_debug3/ 
+
+
+## bbox
+They are ordered as ordered as [xmin,ymin,xmax,ymax] . Cause the coordinates starts from (0,0) of the image which is upper left corner 
+
+xmin ymim - upper left corner
+xmax ymax - bottom lower corner 
+
+![alt text](image-2.png)
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

abstractClass.py

@@ -0,0 +1,59 @@
+
+from typing import Any, List, Literal, Mapping, Optional, Tuple
+from abc import ABC, abstractmethod
+
+import numpy as np
+import cv2
+from PIL import Image
+from abc import ABC, abstractmethod
+
+from utils import cropImage
+
+
+class OCRComponent:
+     """
+     Wrapper class for cropping images and giving it to OCR Predictor 
+     """
+     def predict_pdf(self, pdf_name:str="", page:int=None, bbx:List[List[float]]=None)-> List[List[float]]:
+        #TODO: Preprocessing to crop interest region 
+        pass
+
+
+class TextDetector(ABC):
+    """
+    Abstract base class for text detectors that takes in bounding boxes, pdf name, and page
+    and returns bounding boxes results on them.
+    """
+    
+    def __init__(self):
+        
+        pass
+
+    """
+    This is for predicting given an already cropped image 
+    """
+    @abstractmethod
+    def predict_img(self, img:np.ndarray=None)-> List[List[float]]:
+        # do something with self.input and return bbx 
+        pass
+
+class textRecognizer(ABC):
+    """
+    class of textRecognizer that takes in bounding boxes, pdf name and page and returns 
+    OCR results on them 
+    """
+    
+    def __init__(self):
+        
+        pass
+
+
+    """
+    This is for predicting given text line detection result form text line detector 
+    """
+    @abstractmethod
+    def predict_img(self, bxs:List[List[float]], img:Image.Image)-> List[List[float]]:
+        # do something with self.input and return bbx 
+        pass
+
+   

app.py

@@ -0,0 +1,406 @@
+
+import os
+import traceback
+import argparse
+from typing import List, Tuple, Set, Dict
+
+import time 
+from PIL import Image
+import numpy as np
+from doctr.models import ocr_predictor
+import logging
+import pandas as pd
+from bs4 import BeautifulSoup
+import gradio
+
+from utils import cropImages
+from utils import draw_only_box,draw_box_with_text,getlogger,Annotation
+from ocr_component1 import OCRComponent1
+from detectionAndOcrTable1 import DetectionAndOcrTable1
+from detectionAndOcrTable2 import DetectionAndOcrTable2
+from detectionAndOcrTable3 import DetectionAndOcrTable3
+from detectionAndOcrTable4 import DetectionAndOcrTable4
+from ocrTable1 import OcrTable1
+from ocrTable2 import OcrTable2 
+from pdf2image import convert_from_path
+
+
+def convertHTMLToCSV(html:str,output_path:str)->str:
+    
+    # empty list
+    data = []
+    
+    # for getting the header from
+    # the HTML file
+    list_header = []
+    soup = BeautifulSoup(html,'html.parser')
+    header = soup.find_all("table")[0].find("tr")
+    
+    for items in header:
+        try:
+            list_header.append(items.get_text())
+        except:
+            continue
+    
+    # for getting the data 
+    HTML_data = soup.find_all("table")[0].find_all("tr")[1:]
+    
+    for element in HTML_data:
+        sub_data = []
+        for sub_element in element:
+            try:
+                sub_data.append(sub_element.get_text())
+            except:
+                continue
+        data.append(sub_data)
+    
+    # Storing the data into Pandas
+    # DataFrame 
+    dataFrame = pd.DataFrame(data = data, columns = list_header)
+    
+    # Converting Pandas DataFrame
+    # into CSV file
+    dataFrame.to_csv(output_path)
+
+def saveResults(image_list, results, labels, output_dir='output/', threshold=0.5):
+    if not os.path.exists(output_dir):
+        os.makedirs(output_dir)
+    for idx, im in enumerate(image_list):
+        im = draw_only_box(im, results[idx], labels, threshold=threshold)
+
+        out_path = os.path.join(output_dir, f"{idx}.jpg")
+        im.save(out_path, quality=95)
+        print("save result to: " + out_path)
+
+def InputToImages(input_path:str,resolution=300)-> List[Image.Image]:
+    """
+    input is file location to image 
+    return : List of Pillow image objects
+    """
+    images=[]
+    try:
+        img =Image.open(input_path)
+        if img.mode == 'RGBA':
+            img = img.convert('RGB')
+        images.append(img)
+    except Exception as e:
+        traceback.print_exc()
+    return images
+
+def drawTextDetRes(bxs :List[List[float]],img:Image.Image,output_path:str):
+    """
+    draw layout analysis results 
+    """
+    """bxs_draw is xmin, ymin, xmax, ymax"""
+    bxs_draw = [[b[0][0], b[0][1], b[1][0], b[-1][1]] for b in bxs if b[0][0] <= b[1][0] and b[0][1] <= b[-1][1]]
+    
+    #images_to_recognizer = cropImage(bxs, img)
+    img_to_save = draw_only_box(img, bxs_draw)
+    img_to_save.save(output_path, quality=95)
+
+def test_ocr_component1(test_file="TestingFiles/OCRTest1German.pdf", debug_folder = './res/table1/',englishFlag = False):
+    #Takes as input image of a single page and returns the detected lines and words
+
+    images = convert_from_path(test_file)
+    ocr = OCRComponent1(englishFlag)
+    ocr_results = {}
+
+    all_text_in_pages = {}
+    for page_number,img in enumerate(images):
+        text_in_page = ""
+        
+        line_annotations= ocr.predict(img = np.array(img))
+        ocr_results[page_number] = line_annotations
+
+        """
+        boxes_to_draw =[]
+        for list_of_ann in word_annotations:
+            for ann in list_of_ann:
+                logger.info(ann.text)
+                b = ann.box
+                boxes_to_draw.append(b)
+
+        img_to_save = draw_only_box(img,boxes_to_draw)
+        img_to_save.save("res/12June_2_lines.png", quality=95)
+        """
+
+        line_boxes_to_draw =[]
+        #print("Detected lines are ")
+        #print(len(line_annotations.items()))
+        for index,ann in line_annotations.items():
+
+            b = ann.box
+            line_boxes_to_draw.append(b)
+            line_words = ""
+            #print("detected words per line")
+            #print(len(ann.words))
+            for wordann in ann.words:
+                line_words += wordann.text +" "
+            print(line_words)
+            text_in_page += line_words +"\n"
+                
+        img_to_save1 = draw_only_box(img,line_boxes_to_draw)
+        imgname = test_file.split("/")[-1][:-4]
+        img_to_save1.save(debug_folder+imgname+"_"+str(page_number)+"_bbox_detection.png", quality=95)
+
+        all_text_in_pages[page_number] = text_in_page
+
+    return ocr_results, all_text_in_pages
+
+
+def test_tableOcrOnly1(test_file :Image.Image , debug_folder = './res/table1/',denoise = False,englishFlag = False):
+    #Hybrid Unitable +DocTR
+    #Good at these kind of tables - with a lot of texts 
+    table = OcrTable1(englishFlag)
+    image = test_file.convert("RGB")
+    """
+    parts = test_file.split("/")
+    filename = parts[-1][:-4]
+    debugfolder_filename_page_name= debug_folder+filename+"_"
+    
+    table_code = table.predict([image],debugfolder_filename_page_name,denoise = denoise)
+    with open(debugfolder_filename_page_name+'output.txt', 'w') as file:
+        file.write(table_code)
+    """
+
+    table_code = table.predict([image],denoise = denoise)
+    return table_code
+
+
+def test_tableOcrOnly2(test_file:Image.Image, debug_folder = './res/table2/'):
+    table = OcrTable2()
+    #FullUnitable
+    #Good at these kind of tables - with not much text 
+ 
+    image = test_file.convert("RGB")
+    table.predict([image],debug_folder)
+
+def test_table_component1(test_file = 'TestingFiles/TableOCRTestEnglish.pdf', debug_folder ='./res/table_debug2/',denoise = False,englishFlag = True):
+    table_predictor = DetectionAndOcrTable1(englishFlag)
+
+    images = convert_from_path(test_file)
+    for page_number,img in enumerate(images):
+
+        #print(img.mode)
+        print("Looking at page:")
+        print(page_number)
+        parts = test_file.split("/")
+        filename = parts[-1][:-4]
+        debugfolder_filename_page_name= debug_folder+filename+"_"+ str(page_number)+'_'
+        table_codes = table_predictor.predict(img,debugfolder_filename_page_name=debugfolder_filename_page_name,denoise = denoise)
+        for index, table_code in enumerate(table_codes):
+            with open(debugfolder_filename_page_name+str(index)+'output.xls', 'w') as file:
+                file.write(table_code)
+        return table_codes
+
+def test_table_component2(test_file = 'TestingFiles/TableOCRTestEnglish.pdf', debug_folder ='./res/table_debug2/'):
+    #This components can take in entire pdf page as input , scan for tables and return the table in html format
+    #Uses the full unitable model
+
+    table_predictor = DetectionAndOcrTable2()
+
+    images = convert_from_path(test_file)
+    for page_number,img in enumerate(images):
+        print("Looking at page:")
+        print(page_number)
+        parts = test_file.split("/")
+        filename = parts[-1][:-4]
+        debugfolder_filename_page_name= debug_folder+filename+"_"+ str(page_number)+'_'
+        table_codes = table_predictor.predict(img,debugfolder_filename_page_name=debugfolder_filename_page_name)
+        for index, table_code in enumerate(table_codes):
+            with open(debugfolder_filename_page_name+str(index)+'output.xls', 'w') as file:
+                file.write(table_code)
+        return table_codes
+
+def test_table_component3(test_file = 'TestingFiles/TableOCRTestEnglish.pdf',debug_folder ='./res/table_debug3/',denoise = False,englishFlag = True):
+    table_predictor = DetectionAndOcrTable3(englishFlag)
+
+    images = convert_from_path(test_file)
+    for page_number,img in enumerate(images):
+        #print(img.mode)
+        print("Looking at page:")
+        print(page_number)
+        parts = test_file.split("/")
+        filename = parts[-1][:-4]
+        debugfolder_filename_page_name= debug_folder+filename+"_"+ str(page_number)+'_'
+        table_codes = table_predictor.predict(img,debugfolder_filename_page_name=debugfolder_filename_page_name)
+        for index, table_code in enumerate(table_codes):
+            with open(debugfolder_filename_page_name+str(index)+'output.xls', 'w') as file:
+                file.write(table_code)
+        return table_codes
+
+
+
+def test_table_component4(test_file = 'TestingFiles/TableOCRTestEnglish.pdf',debug_folder ='./res/table_debug3/'):
+    table_predictor = DetectionAndOcrTable4()
+
+    images = convert_from_path(test_file)
+    for page_number,img in enumerate(images):
+        #print(img.mode)
+        print("Looking at page:")
+        print(page_number)
+        parts = test_file.split("/")
+        filename = parts[-1][:-4]
+        debugfolder_filename_page_name= debug_folder+filename+"_"+ str(page_number)+'_'
+        table_codes = table_predictor.predict(img,debugfolder_filename_page_name=debugfolder_filename_page_name)
+        for index, table_code in enumerate(table_codes):
+            with open(debugfolder_filename_page_name+str(index)+'output.xls', 'w') as file:
+                file.write(table_code)
+        return table_codes
+
+
+"""
+    parser = argparse.ArgumentParser(description='Process some strings.')
+    parser.add_argument('ocr', type=str, help='type in id of the component to test')
+    parser.add_argument('--test_file',type=str, help='path to the testing file')
+    parser.add_argument('--debug_folder',type=str, help='path to the folder you want to save your results in')
+    parser.add_argument('--englishFlag',type=bool, help='Whether your pdf is in english => could lead to better results ')
+    parser.add_argument('--denoise',type=bool, help='preprocessing for not clean scans  ')
+
+    args = parser.parse_args()
+    start = time.time()
+    if args.ocr == "ocr1":
+        test_ocr_component1(args.test_file,args.debug_folder, args.englishFlag)
+    elif args.ocr == "table1":
+        test_tableOcrOnly1(args.test_file,args.debug_folder,args.englishFlag,args.denoise)
+    elif args.ocr == "table2":
+        test_tableOcrOnly2(args.test_file,args.debug_folder)
+    elif args.ocr =="pdftable1":
+        test_table_component1(args.test_file,args.debug_folder,args.englishFlag,args.denoise)
+    elif args.ocr =="pdftable2":
+        test_table_component2(args.test_file,args.debug_folder)
+    elif args.ocr =="pdftable3":
+        test_table_component3(args.test_file,args.debug_folder,args.englishFlag,args.denoise)
+    elif args.ocr =="pdftable4":
+        test_table_component4(args.test_file,args.debug_folder)
+
+"""
+import gradio as gr
+from gradio_pdf import PDF
+
+with gr.Blocks() as demo:
+    gr.Markdown("# OCR component")
+    inputs_for_ocr = [PDF(label="Document"), gr.Textbox(label="internal debug folder",placeholder = "./res/table1/"),gr.Checkbox(label ="English Document?",value =False)]
+    ocr_btn = gr.Button("Run ocr")
+   
+    gr.Examples(
+        examples=[["TestingFiles/OCRTest1German.pdf",'./res/table1/',False]],
+        inputs=inputs_for_ocr
+                )
+    
+    outputs_for_ocr = [gr.Textbox(label="List of annotation objects"), gr.Textbox("Text in page")]
+
+    ocr_btn.click(fn=test_ocr_component1, 
+                  inputs = inputs_for_ocr,
+                  outputs = outputs_for_ocr,
+                  api_name="OCR"
+                  )
+    
+    gr.Markdown("# Table OCR components that takes a pdf, extract table and return their html code ")
+    gr.Markdown("## Component 1 uses table transformer and doctr +Unitable")
+    inputs_for_pdftable1 = [PDF(label="Document"), gr.Textbox(label="internal debug folder",placeholder = "./res/table1/"),gr.Checkbox(label ="Denoise?",value =False),gr.Checkbox(label ="English Document?",value =False)]
+    table1_btn = gr.Button("Run pdftable1")
+
+    gr.Examples(
+        examples=[["TestingFiles/OCRTest5English.pdf",'./res/table1/',False]],
+        inputs=inputs_for_pdftable1
+                )
+    outputs_for_pdftable1 = [gr.Textbox(label="Table code")]
+    
+    table1_btn.click(fn=test_table_component1,
+                    inputs = inputs_for_pdftable1,
+                    outputs = outputs_for_pdftable1,
+                    api_name="pdfTable1"
+                    )
+    
+    gr.Markdown("## Component 2 uses table transformer and Unitable")
+    inputs_for_pdftable2 = [PDF(label="Document"), gr.Textbox(label="internal debug folder",placeholder = "./res/table1/")]
+    table2_btn = gr.Button("Run pdftable2")
+
+    gr.Examples(
+        examples=[["TestingFiles/OCRTest5English.pdf",'./res/table1/',False]],
+        inputs=inputs_for_pdftable1
+                )
+    outputs_for_pdftable2 = [gr.Textbox(label="Table code")]
+
+    table2_btn.click(fn=test_table_component2,
+                    inputs = inputs_for_pdftable2,
+                    outputs = outputs_for_pdftable2,
+                    api_name="pdfTable2"
+                    )
+    
+    gr.Markdown("## Component 3 uses Yolo and Unitable+doctr")
+    inputs_for_pdftable3 = [PDF(label="Document"), gr.Textbox(label="internal debug folder",placeholder = "./res/table1/"),gr.Checkbox(label ="Denoise?",value =False),gr.Checkbox(label ="English Document?",value =False)]
+    table3_btn = gr.Button("Run pdftable3")
+    
+
+    gr.Examples(
+        examples=[["TestingFiles/TableOCRTestEnglish.pdf",'./res/table1/',False]],
+        inputs=inputs_for_pdftable1
+                )
+    outputs_for_pdftable3 = [gr.Textbox(label="Table code")]
+
+    table3_btn.click(fn=test_table_component3,
+                    inputs = inputs_for_pdftable3,
+                    outputs = outputs_for_pdftable3,
+                    api_name="pdfTable3"
+                    )
+    
+    gr.Markdown("## Component 4 uses Yolo and Unitable")
+    inputs_for_pdftable4 = [PDF(label="Document"), gr.Textbox(label="internal debug folder",placeholder = "./res/table1/")]
+    table4_btn = gr.Button("Run pdftable4")
+    
+    gr.Examples(
+        examples=[["TestingFiles/TableOCRTestEasier.pdf",'./res/table1/',False]],
+        inputs=inputs_for_pdftable1
+                )
+    outputs_for_pdftable4 = [gr.Textbox(label="Table code")]
+
+
+    table4_btn.click(fn=test_table_component4,
+                    inputs = inputs_for_pdftable4,
+                    outputs = outputs_for_pdftable4,
+                    api_name="pdfTable4"
+                    )
+    
+
+    gr.Markdown("# Table OCR component that takes image of an cropped tavle, extract table and return their html code ")
+
+    inputs_for_table1 = [gr.Image(label="Image of cropped table",type='pil'), gr.Textbox(label="internal debug folder",placeholder = "./res/table1/"),gr.Checkbox(label ="Denoise?",value =False),gr.Checkbox(label ="English Document?",value =False)]
+    onlytable1_btn = gr.Button("Run table1")
+    
+    gr.Examples(
+        examples=[[Image.open("cropped_table.png"),'./res/table1/',False]],
+        inputs=inputs_for_table1
+                )
+    outputs_for_table1 = [gr.HTML(label="Table code")]
+
+
+    onlytable1_btn.click(fn=test_tableOcrOnly1,
+                    inputs = inputs_for_table1,
+                    outputs = outputs_for_table1,
+                    api_name="table1"
+                    )
+    
+    gr.Markdown("## Another Table OCR component that takes image of an cropped table, extract table and return their html code ")
+
+    inputs_for_table2 = [gr.Image(label="Image of cropped table",type='pil'), gr.Textbox(label="internal debug folder",placeholder = "./res/table1/")]
+    onlytable2_btn = gr.Button("Run table2")
+    
+    
+    gr.Examples(
+        examples=[[Image.open("cropped_table.png"),'./res/table1/',False]],
+        inputs=inputs_for_table2
+                )
+    outputs_for_table2 = [gr.HTML(label="Table code")]
+    
+    onlytable2_btn.click(fn=test_tableOcrOnly2,
+                    inputs = inputs_for_table2,
+                    outputs = outputs_for_table2,
+                    api_name="table2"
+                    )
+
+
+    
+
+demo.launch(share=True)

deepdoc/README.md

@@ -0,0 +1,122 @@
+English | [简体中文](./README_zh.md)
+
+# *Deep*Doc
+
+- [1. Introduction](#1)
+- [2. Vision](#2)
+- [3. Parser](#3)
+
+<a name="1"></a>
+## 1. Introduction
+
+With a bunch of documents from various domains with various formats and along with diverse retrieval requirements, 
+an accurate analysis becomes a very challenge task. *Deep*Doc is born for that purpose.
+There are 2 parts in *Deep*Doc so far: vision and parser. 
+You can run the flowing test programs if you're interested in our results of OCR, layout recognition and TSR.
+```bash
+python deepdoc/vision/t_ocr.py -h
+usage: t_ocr.py [-h] --inputs INPUTS [--output_dir OUTPUT_DIR]
+
+options:
+  -h, --help            show this help message and exit
+  --inputs INPUTS       Directory where to store images or PDFs, or a file path to a single image or PDF
+  --output_dir OUTPUT_DIR
+                        Directory where to store the output images. Default: './ocr_outputs'
+```
+```bash
+python deepdoc/vision/t_recognizer.py -h
+usage: t_recognizer.py [-h] --inputs INPUTS [--output_dir OUTPUT_DIR] [--threshold THRESHOLD] [--mode {layout,tsr}]
+
+options:
+  -h, --help            show this help message and exit
+  --inputs INPUTS       Directory where to store images or PDFs, or a file path to a single image or PDF
+  --output_dir OUTPUT_DIR
+                        Directory where to store the output images. Default: './layouts_outputs'
+  --threshold THRESHOLD
+                        A threshold to filter out detections. Default: 0.5
+  --mode {layout,tsr}   Task mode: layout recognition or table structure recognition
+```
+
+Our models are served on HuggingFace. If you have trouble downloading HuggingFace models, this might help!!
+```bash
+export HF_ENDPOINT=https://hf-mirror.com
+```
+
+<a name="2"></a>
+## 2. Vision
+
+We use vision information to resolve problems as human being.
+  - OCR. Since a lot of documents presented as images or at least be able to transform to image, 
+    OCR is a very essential and fundamental or even universal solution for text extraction.
+    ```bash
+        python deepdoc/vision/t_ocr.py --inputs=path_to_images_or_pdfs --output_dir=path_to_store_result
+     ```
+    The inputs could be directory to images or PDF, or a image or PDF. 
+    You can look into the folder 'path_to_store_result' where has images which demonstrate the positions of results,
+    txt files which contain the OCR text.
+    <div align="center" style="margin-top:20px;margin-bottom:20px;">
+    <img src="https://github.com/infiniflow/ragflow/assets/12318111/f25bee3d-aaf7-4102-baf5-d5208361d110" width="900"/>
+    </div>
+
+  - Layout recognition. Documents from different domain may have various layouts, 
+    like, newspaper, magazine, book and résumé are distinct in terms of layout. 
+    Only when machine have an accurate layout analysis, it can decide if these text parts are successive or not, 
+    or this part needs Table Structure Recognition(TSR) to process, or this part is a figure and described with this caption.
+    We have 10 basic layout components which covers most cases:
+      - Text
+      - Title
+      - Figure
+      - Figure caption
+      - Table
+      - Table caption
+      - Header
+      - Footer
+      - Reference
+      - Equation
+      
+     Have a try on the following command to see the layout detection results.
+     ```bash
+        python deepdoc/vision/t_recognizer.py --inputs=path_to_images_or_pdfs --threshold=0.2 --mode=layout --output_dir=path_to_store_result
+     ```
+    The inputs could be directory to images or PDF, or a image or PDF. 
+    You can look into the folder 'path_to_store_result' where has images which demonstrate the detection results as following:
+    <div align="center" style="margin-top:20px;margin-bottom:20px;">
+    <img src="https://github.com/infiniflow/ragflow/assets/12318111/07e0f625-9b28-43d0-9fbb-5bf586cd286f" width="1000"/>
+    </div>
+  
+  - Table Structure Recognition(TSR). Data table is a frequently used structure to present data including numbers or text.
+    And the structure of a table might be very complex, like hierarchy headers, spanning cells and projected row headers.
+    Along with TSR, we also reassemble the content into sentences which could be well comprehended by LLM. 
+    We have five labels for TSR task:
+      - Column
+      - Row
+      - Column header
+      - Projected row header
+      - Spanning cell
+      
+    Have a try on the following command to see the layout detection results.
+     ```bash
+        python deepdoc/vision/t_recognizer.py --inputs=path_to_images_or_pdfs --threshold=0.2 --mode=tsr --output_dir=path_to_store_result
+     ```
+    The inputs could be directory to images or PDF, or a image or PDF. 
+    You can look into the folder 'path_to_store_result' where has both images and html pages which demonstrate the detection results as following:
+    <div align="center" style="margin-top:20px;margin-bottom:20px;">
+    <img src="https://github.com/infiniflow/ragflow/assets/12318111/cb24e81b-f2ba-49f3-ac09-883d75606f4c" width="1000"/>
+    </div>
+        
+<a name="3"></a>
+## 3. Parser
+
+Four kinds of document formats as PDF, DOCX, EXCEL and PPT have their corresponding parser. 
+The most complex one is PDF parser since PDF's flexibility. The output of PDF parser includes:
+  - Text chunks with their own positions in PDF(page number and rectangular positions).
+  - Tables with cropped image from the PDF, and contents which has already translated into natural language sentences.
+  - Figures with caption and text in the figures.
+  
+### Résumé
+
+The résumé is a very complicated kind of document. A résumé which is composed of unstructured text 
+with various layouts could be resolved into structured data composed of nearly a hundred of fields.
+We haven't opened the parser yet, as we open the processing method after parsing procedure.
+
+    

deepdoc/__init__.py

@@ -0,0 +1,8 @@
+
+"""
+In deepdoc/__init__.py, import the class from ragFlow.py and make it available for import from the deepdoc package:
+"""
+
+from .vision import RagFlow
+
+__all__ = ['RagFlow']

deepdoc/models/.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

deepdoc/models/README.md

@@ -0,0 +1,3 @@
+---
+license: apache-2.0
+---

deepdoc/models/det.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30a86f5731181461d08021402766601e4302a9b9b9666be8aff402696339cdff
+size 4745517

deepdoc/models/layout.laws.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:485a7ddf6889ef15a150bded7091ec1ea5467871f50a88f5f4297c66c1ecef1e
+size 12246134

deepdoc/models/layout.manual.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:485a7ddf6889ef15a150bded7091ec1ea5467871f50a88f5f4297c66c1ecef1e
+size 12246134

deepdoc/models/layout.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:485a7ddf6889ef15a150bded7091ec1ea5467871f50a88f5f4297c66c1ecef1e
+size 12246134

deepdoc/models/layout.paper.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:485a7ddf6889ef15a150bded7091ec1ea5467871f50a88f5f4297c66c1ecef1e
+size 12246134

deepdoc/models/ocr.res

@@ -0,0 +1,6623 @@
+'
+疗
+绚
+诚
+娇
+溜
+题
+贿
+者
+廖
+更
+纳
+加
+奉
+公
+一
+就
+汴
+计
+与
+路
+房
+原
+妇
+2
+0
+8
+-
+7
+其
+>
+:
+]
+,
+，
+骑
+刈
+全
+消
+昏
+傈
+安
+久
+钟
+嗅
+不
+影
+处
+驽
+蜿
+资
+关
+椤
+地
+瘸
+专
+问
+忖
+票
+嫉
+炎
+韵
+要
+月
+田
+节
+陂
+鄙
+捌
+备
+拳
+伺
+眼
+网
+盎
+大
+傍
+心
+东
+愉
+汇
+蹿
+科
+每
+业
+里
+航
+晏
+字
+平
+录
+先
+1
+3
+彤
+鲶
+产
+稍
+督
+腴
+有
+象
+岳
+注
+绍
+在
+泺
+文
+定
+核
+名
+水
+过
+理
+让
+偷
+率
+等
+这
+发
+”
+为
+含
+肥
+酉
+相
+鄱
+七
+编
+猥
+锛
+日
+镀
+蒂
+掰
+倒
+辆
+栾
+栗
+综
+涩
+州
+雌
+滑
+馀
+了
+机
+块
+司
+宰
+甙
+兴
+矽
+抚
+保
+用
+沧
+秩
+如
+收
+息
+滥
+页
+疑
+埠
+!
+！
+姥
+异
+橹
+钇
+向
+下
+跄
+的
+椴
+沫
+国
+绥
+獠
+报
+开
+民
+蜇
+何
+分
+凇
+长
+讥
+藏
+掏
+施
+羽
+中
+讲
+派
+嘟
+人
+提
+浼
+间
+世
+而
+古
+多
+倪
+唇
+饯
+控
+庚
+首
+赛
+蜓
+味
+断
+制
+觉
+技
+替
+艰
+溢
+潮
+夕
+钺
+外
+摘
+枋
+动
+双
+单
+啮
+户
+枇
+确
+锦
+曜
+杜
+或
+能
+效
+霜
+盒
+然
+侗
+电
+晁
+放
+步
+鹃
+新
+杖
+蜂
+吒
+濂
+瞬
+评
+总
+隍
+对
+独
+合
+也
+是
+府
+青
+天
+诲
+墙
+组
+滴
+级
+邀
+帘
+示
+已
+时
+骸
+仄
+泅
+和
+遨
+店
+雇
+疫
+持
+巍
+踮
+境
+只
+亨
+目
+鉴
+崤
+闲
+体
+泄
+杂
+作
+般
+轰
+化
+解
+迂
+诿
+蛭
+璀
+腾
+告
+版
+服
+省
+师
+小
+规
+程
+线
+海
+办
+引
+二
+桧
+牌
+砺
+洄
+裴
+修
+图
+痫
+胡
+许
+犊
+事
+郛
+基
+柴
+呼
+食
+研
+奶
+律
+蛋
+因
+葆
+察
+戏
+褒
+戒
+再
+李
+骁
+工
+貂
+油
+鹅
+章
+啄
+休
+场
+给
+睡
+纷
+豆
+器
+捎
+说
+敏
+学
+会
+浒
+设
+诊
+格
+廓
+查
+来
+霓
+室
+溆
+￠
+诡
+寥
+焕
+舜
+柒
+狐
+回
+戟
+砾
+厄
+实
+翩
+尿
+五
+入
+径
+惭
+喹
+股
+宇
+篝
+|
+;
+美
+期
+云
+九
+祺
+扮
+靠
+锝
+槌
+系
+企
+酰
+阊
+暂
+蚕
+忻
+豁
+本
+羹
+执
+条
+钦
+H
+獒
+限
+进
+季
+楦
+于
+芘
+玖
+铋
+茯
+未
+答
+粘
+括
+样
+精
+欠
+矢
+甥
+帷
+嵩
+扣
+令
+仔
+风
+皈
+行
+支
+部
+蓉
+刮
+站
+蜡
+救
+钊
+汗
+松
+嫌
+成
+可
+.
+鹤
+院
+从
+交
+政
+怕
+活
+调
+球
+局
+验
+髌
+第
+韫
+谗
+串
+到
+圆
+年
+米
+/
+*
+友
+忿
+检
+区
+看
+自
+敢
+刃
+个
+兹
+弄
+流
+留
+同
+没
+齿
+星
+聆
+轼
+湖
+什
+三
+建
+蛔
+儿
+椋
+汕
+震
+颧
+鲤
+跟
+力
+情
+璺
+铨
+陪
+务
+指
+族
+训
+滦
+鄣
+濮
+扒
+商
+箱
+十
+召
+慷
+辗
+所
+莞
+管
+护
+臭
+横
+硒
+嗓
+接
+侦
+六
+露
+党
+馋
+驾
+剖
+高
+侬
+妪
+幂
+猗
+绺
+骐
+央
+酐
+孝
+筝
+课
+徇
+缰
+门
+男
+西
+项
+句
+谙
+瞒
+秃
+篇
+教
+碲
+罚
+声
+呐
+景
+前
+富
+嘴
+鳌
+稀
+免
+朋
+啬
+睐
+去
+赈
+鱼
+住
+肩
+愕
+速
+旁
+波
+厅
+健
+茼
+厥
+鲟
+谅
+投
+攸
+炔
+数
+方
+击
+呋
+谈
+绩
+别
+愫
+僚
+躬
+鹧
+胪
+炳
+招
+喇
+膨
+泵
+蹦
+毛
+结
+5
+4
+谱
+识
+陕
+粽
+婚
+拟
+构
+且
+搜
+任
+潘
+比
+郢
+妨
+醪
+陀
+桔
+碘
+扎
+选
+哈
+骷
+楷
+亿
+明
+缆
+脯
+监
+睫
+逻
+婵
+共
+赴
+淝
+凡
+惦
+及
+达
+揖
+谩
+澹
+减
+焰
+蛹
+番
+祁
+柏
+员
+禄
+怡
+峤
+龙
+白
+叽
+生
+闯
+起
+细
+装
+谕
+竟
+聚
+钙
+上
+导
+渊
+按
+艾
+辘
+挡
+耒
+盹
+饪
+臀
+记
+邮
+蕙
+受
+各
+医
+搂
+普
+滇
+朗
+茸
+带
+翻
+酚
+(
+光
+堤
+墟
+蔷
+万
+幻
+〓
+瑙
+辈
+昧
+盏
+亘
+蛀
+吉
+铰
+请
+子
+假
+闻
+税
+井
+诩
+哨
+嫂
+好
+面
+琐
+校
+馊
+鬣
+缂
+营
+访
+炖
+占
+农
+缀
+否
+经
+钚
+棵
+趟
+张
+亟
+吏
+茶
+谨
+捻
+论
+迸
+堂
+玉
+信
+吧
+瞠
+乡
+姬
+寺
+咬
+溏
+苄
+皿
+意
+赉
+宝
+尔
+钰
+艺
+特
+唳
+踉
+都
+荣
+倚
+登
+荐
+丧
+奇
+涵
+批
+炭
+近
+符
+傩
+感
+道
+着
+菊
+虹
+仲
+众
+懈
+濯
+颞
+眺
+南
+释
+北
+缝
+标
+既
+茗
+整
+撼
+迤
+贲
+挎
+耱
+拒
+某
+妍
+卫
+哇
+英
+矶
+藩
+治
+他
+元
+领
+膜
+遮
+穗
+蛾
+飞
+荒
+棺
+劫
+么
+市
+火
+温
+拈
+棚
+洼
+转
+果
+奕
+卸
+迪
+伸
+泳
+斗
+邡
+侄
+涨
+屯
+萋
+胭
+氡
+崮
+枞
+惧
+冒
+彩
+斜
+手
+豚
+随
+旭
+淑
+妞
+形
+菌
+吲
+沱
+争
+驯
+歹
+挟
+兆
+柱
+传
+至
+包
+内
+响
+临
+红
+功
+弩
+衡
+寂
+禁
+老
+棍
+耆
+渍
+织
+害
+氵
+渑
+布
+载
+靥
+嗬
+虽
+苹
+咨
+娄
+库
+雉
+榜
+帜
+嘲
+套
+瑚
+亲
+簸
+欧
+边
+6
+腿
+旮
+抛
+吹
+瞳
+得
+镓
+梗
+厨
+继
+漾
+愣
+憨
+士
+策
+窑
+抑
+躯
+襟
+脏
+参
+贸
+言
+干
+绸
+鳄
+穷
+藜
+音
+折
+详
+)
+举
+悍
+甸
+癌
+黎
+谴
+死
+罩
+迁
+寒
+驷
+袖
+媒
+蒋
+掘
+模
+纠
+恣
+观
+祖
+蛆
+碍
+位
+稿
+主
+澧
+跌
+筏
+京
+锏
+帝
+贴
+证
+糠
+才
+黄
+鲸
+略
+炯
+饱
+四
+出
+园
+犀
+牧
+容
+汉
+杆
+浈
+汰
+瑷
+造
+虫
+瘩
+怪
+驴
+济
+应
+花
+沣
+谔
+夙
+旅
+价
+矿
+以
+考
+s
+u
+呦
+晒
+巡
+茅
+准
+肟
+瓴
+詹
+仟
+褂
+译
+桌
+混
+宁
+怦
+郑
+抿
+些
+余
+鄂
+饴
+攒
+珑
+群
+阖
+岔
+琨
+藓
+预
+环
+洮
+岌
+宀
+杲
+瀵
+最
+常
+囡
+周
+踊
+女
+鼓
+袭
+喉
+简
+范
+薯
+遐
+疏
+粱
+黜
+禧
+法
+箔
+斤
+遥
+汝
+奥
+直
+贞
+撑
+置
+绱
+集
+她
+馅
+逗
+钧
+橱
+魉
+[
+恙
+躁
+唤
+9
+旺
+膘
+待
+脾
+惫
+购
+吗
+依
+盲
+度
+瘿
+蠖
+俾
+之
+镗
+拇
+鲵
+厝
+簧
+续
+款
+展
+啃
+表
+剔
+品
+钻
+腭
+损
+清
+锶
+统
+涌
+寸
+滨
+贪
+链
+吠
+冈
+伎
+迥
+咏
+吁
+览
+防
+迅
+失
+汾
+阔
+逵
+绀
+蔑
+列
+川
+凭
+努
+熨
+揪
+利
+俱
+绉
+抢
+鸨
+我
+即
+责
+膦
+易
+毓
+鹊
+刹
+玷
+岿
+空
+嘞
+绊
+排
+术
+估
+锷
+违
+们
+苟
+铜
+播
+肘
+件
+烫
+审
+鲂
+广
+像
+铌
+惰
+铟
+巳
+胍
+鲍
+康
+憧
+色
+恢
+想
+拷
+尤
+疳
+知
+S
+Y
+F
+D
+A
+峄
+裕
+帮
+握
+搔
+氐
+氘
+难
+墒
+沮
+雨
+叁
+缥
+悴
+藐
+湫
+娟
+苑
+稠
+颛
+簇
+后
+阕
+闭
+蕤
+缚
+怎
+佞
+码
+嘤
+蔡
+痊
+舱
+螯
+帕
+赫
+昵
+升
+烬
+岫
+、
+疵
+蜻
+髁
+蕨
+隶
+烛
+械
+丑
+盂
+梁
+强
+鲛
+由
+拘
+揉
+劭
+龟
+撤
+钩
+呕
+孛
+费
+妻
+漂
+求
+阑
+崖
+秤
+甘
+通
+深
+补
+赃
+坎
+床
+啪
+承
+吼
+量
+暇
+钼
+烨
+阂
+擎
+脱
+逮
+称
+P
+神
+属
+矗
+华
+届
+狍
+葑
+汹
+育
+患
+窒
+蛰
+佼
+静
+槎
+运
+鳗
+庆
+逝
+曼
+疱
+克
+代
+官
+此
+麸
+耧
+蚌
+晟
+例
+础
+榛
+副
+测
+唰
+缢
+迹
+灬
+霁
+身
+岁
+赭
+扛
+又
+菡
+乜
+雾
+板
+读
+陷
+徉
+贯
+郁
+虑
+变
+钓
+菜
+圾
+现
+琢
+式
+乐
+维
+渔
+浜
+左
+吾
+脑
+钡
+警
+T
+啵
+拴
+偌
+漱
+湿
+硕
+止
+骼
+魄
+积
+燥
+联
+踢
+玛
+则
+窿
+见
+振
+畿
+送
+班
+钽
+您
+赵
+刨
+印
+讨
+踝
+籍
+谡
+舌
+崧
+汽
+蔽
+沪
+酥
+绒
+怖
+财
+帖
+肱
+私
+莎
+勋
+羔
+霸
+励
+哼
+帐
+将
+帅
+渠
+纪
+婴
+娩
+岭
+厘
+滕
+吻
+伤
+坝
+冠
+戊
+隆
+瘁
+介
+涧
+物
+黍
+并
+姗
+奢
+蹑
+掣
+垸
+锴
+命
+箍
+捉
+病
+辖
+琰
+眭
+迩
+艘
+绌
+繁
+寅
+若
+毋
+思
+诉
+类
+诈
+燮
+轲
+酮
+狂
+重
+反
+职
+筱
+县
+委
+磕
+绣
+奖
+晋
+濉
+志
+徽
+肠
+呈
+獐
+坻
+口
+片
+碰
+几
+村
+柿
+劳
+料
+获
+亩
+惕
+晕
+厌
+号
+罢
+池
+正
+鏖
+煨
+家
+棕
+复
+尝
+懋
+蜥
+锅
+岛
+扰
+队
+坠
+瘾
+钬
+@
+卧
+疣
+镇
+譬
+冰
+彷
+频
+黯
+据
+垄
+采
+八
+缪
+瘫
+型
+熹
+砰
+楠
+襁
+箐
+但
+嘶
+绳
+啤
+拍
+盥
+穆
+傲
+洗
+盯
+塘
+怔
+筛
+丿
+台
+恒
+喂
+葛
+永
+￥
+烟
+酒
+桦
+书
+砂
+蚝
+缉
+态
+瀚
+袄
+圳
+轻
+蛛
+超
+榧
+遛
+姒
+奘
+铮
+右
+荽
+望
+偻
+卡
+丶
+氰
+附
+做
+革
+索
+戚
+坨
+桷
+唁
+垅
+榻
+岐
+偎
+坛
+莨
+山
+殊
+微
+骇
+陈
+爨
+推
+嗝
+驹
+澡
+藁
+呤
+卤
+嘻
+糅
+逛
+侵
+郓
+酌
+德
+摇
+※
+鬃
+被
+慨
+殡
+羸
+昌
+泡
+戛
+鞋
+河
+宪
+沿
+玲
+鲨
+翅
+哽
+源
+铅
+语
+照
+邯
+址
+荃
+佬
+顺
+鸳
+町
+霭
+睾
+瓢
+夸
+椁
+晓
+酿
+痈
+咔
+侏
+券
+噎
+湍
+签
+嚷
+离
+午
+尚
+社
+锤
+背
+孟
+使
+浪
+缦
+潍
+鞅
+军
+姹
+驶
+笑
+鳟
+鲁
+》
+孽
+钜
+绿
+洱
+礴
+焯
+椰
+颖
+囔
+乌
+孔
+巴
+互
+性
+椽
+哞
+聘
+昨
+早
+暮
+胶
+炀
+隧
+低
+彗
+昝
+铁
+呓
+氽
+藉
+喔
+癖
+瑗
+姨
+权
+胱
+韦
+堑
+蜜
+酋
+楝
+砝
+毁
+靓
+歙
+锲
+究
+屋
+喳
+骨
+辨
+碑
+武
+鸠
+宫
+辜
+烊
+适
+坡
+殃
+培
+佩
+供
+走
+蜈
+迟
+翼
+况
+姣
+凛
+浔
+吃
+飘
+债
+犟
+金
+促
+苛
+崇
+坂
+莳
+畔
+绂
+兵
+蠕
+斋
+根
+砍
+亢
+欢
+恬
+崔
+剁
+餐
+榫
+快
+扶
+‖
+濒
+缠
+鳜
+当
+彭
+驭
+浦
+篮
+昀
+锆
+秸
+钳
+弋
+娣
+瞑
+夷
+龛
+苫
+拱
+致
+%
+嵊
+障
+隐
+弑
+初
+娓
+抉
+汩
+累
+蓖
+"
+唬
+助
+苓
+昙
+押
+毙
+破
+城
+郧
+逢
+嚏
+獭
+瞻
+溱
+婿
+赊
+跨
+恼
+璧
+萃
+姻
+貉
+灵
+炉
+密
+氛
+陶
+砸
+谬
+衔
+点
+琛
+沛
+枳
+层
+岱
+诺
+脍
+榈
+埂
+征
+冷
+裁
+打
+蹴
+素
+瘘
+逞
+蛐
+聊
+激
+腱
+萘
+踵
+飒
+蓟
+吆
+取
+咙
+簋
+涓
+矩
+曝
+挺
+揣
+座
+你
+史
+舵
+焱
+尘
+苏
+笈
+脚
+溉
+榨
+诵
+樊
+邓
+焊
+义
+庶
+儋
+蟋
+蒲
+赦
+呷
+杞
+诠
+豪
+还
+试
+颓
+茉
+太
+除
+紫
+逃
+痴
+草
+充
+鳕
+珉
+祗
+墨
+渭
+烩
+蘸
+慕
+璇
+镶
+穴
+嵘
+恶
+骂
+险
+绋
+幕
+碉
+肺
+戳
+刘
+潞
+秣
+纾
+潜
+銮
+洛
+须
+罘
+销
+瘪
+汞
+兮
+屉
+r
+林
+厕
+质
+探
+划
+狸
+殚
+善
+煊
+烹
+〒
+锈
+逯
+宸
+辍
+泱
+柚
+袍
+远
+蹋
+嶙
+绝
+峥
+娥
+缍
+雀
+徵
+认
+镱
+谷
+=
+贩
+勉
+撩
+鄯
+斐
+洋
+非
+祚
+泾
+诒
+饿
+撬
+威
+晷
+搭
+芍
+锥
+笺
+蓦
+候
+琊
+档
+礁
+沼
+卵
+荠
+忑
+朝
+凹
+瑞
+头
+仪
+弧
+孵
+畏
+铆
+突
+衲
+车
+浩
+气
+茂
+悖
+厢
+枕
+酝
+戴
+湾
+邹
+飚
+攘
+锂
+写
+宵
+翁
+岷
+无
+喜
+丈
+挑
+嗟
+绛
+殉
+议
+槽
+具
+醇
+淞
+笃
+郴
+阅
+饼
+底
+壕
+砚
+弈
+询
+缕
+庹
+翟
+零
+筷
+暨
+舟
+闺
+甯
+撞
+麂
+茌
+蔼
+很
+珲
+捕
+棠
+角
+阉
+媛
+娲
+诽
+剿
+尉
+爵
+睬
+韩
+诰
+匣
+危
+糍
+镯
+立
+浏
+阳
+少
+盆
+舔
+擘
+匪
+申
+尬
+铣
+旯
+抖
+赘
+瓯
+居
+ˇ
+哮
+游
+锭
+茏
+歌
+坏
+甚
+秒
+舞
+沙
+仗
+劲
+潺
+阿
+燧
+郭
+嗖
+霏
+忠
+材
+奂
+耐
+跺
+砀
+输
+岖
+媳
+氟
+极
+摆
+灿
+今
+扔
+腻
+枝
+奎
+药
+熄
+吨
+话
+q
+额
+慑
+嘌
+协
+喀
+壳
+埭
+视
+著
+於
+愧
+陲
+翌
+峁
+颅
+佛
+腹
+聋
+侯
+咎
+叟
+秀
+颇
+存
+较
+罪
+哄
+岗
+扫
+栏
+钾
+羌
+己
+璨
+枭
+霉
+煌
+涸
+衿
+键
+镝
+益
+岢
+奏
+连
+夯
+睿
+冥
+均
+糖
+狞
+蹊
+稻
+爸
+刿
+胥
+煜
+丽
+肿
+璃
+掸
+跚
+灾
+垂
+樾
+濑
+乎
+莲
+窄
+犹
+撮
+战
+馄
+软
+络
+显
+鸢
+胸
+宾
+妲
+恕
+埔
+蝌
+份
+遇
+巧
+瞟
+粒
+恰
+剥
+桡
+博
+讯
+凯
+堇
+阶
+滤
+卖
+斌
+骚
+彬
+兑
+磺
+樱
+舷
+两
+娱
+福
+仃
+差
+找
+桁
+÷
+净
+把
+阴
+污
+戬
+雷
+碓
+蕲
+楚
+罡
+焖
+抽
+妫
+咒
+仑
+闱
+尽
+邑
+菁
+爱
+贷
+沥
+鞑
+牡
+嗉
+崴
+骤
+塌
+嗦
+订
+拮
+滓
+捡
+锻
+次
+坪
+杩
+臃
+箬
+融
+珂
+鹗
+宗
+枚
+降
+鸬
+妯
+阄
+堰
+盐
+毅
+必
+杨
+崃
+俺
+甬
+状
+莘
+货
+耸
+菱
+腼
+铸
+唏
+痤
+孚
+澳
+懒
+溅
+翘
+疙
+杷
+淼
+缙
+骰
+喊
+悉
+砻
+坷
+艇
+赁
+界
+谤
+纣
+宴
+晃
+茹
+归
+饭
+梢
+铡
+街
+抄
+肼
+鬟
+苯
+颂
+撷
+戈
+炒
+咆
+茭
+瘙
+负
+仰
+客
+琉
+铢
+封
+卑
+珥
+椿
+镧
+窨
+鬲
+寿
+御
+袤
+铃
+萎
+砖
+餮
+脒
+裳
+肪
+孕
+嫣
+馗
+嵇
+恳
+氯
+江
+石
+褶
+冢
+祸
+阻
+狈
+羞
+银
+靳
+透
+咳
+叼
+敷
+芷
+啥
+它
+瓤
+兰
+痘
+懊
+逑
+肌
+往
+捺
+坊
+甩
+呻
+〃
+沦
+忘
+膻
+祟
+菅
+剧
+崆
+智
+坯
+臧
+霍
+墅
+攻
+眯
+倘
+拢
+骠
+铐
+庭
+岙
+瓠
+′
+缺
+泥
+迢
+捶
+?
+？
+郏
+喙
+掷
+沌
+纯
+秘
+种
+听
+绘
+固
+螨
+团
+香
+盗
+妒
+埚
+蓝
+拖
+旱
+荞
+铀
+血
+遏
+汲
+辰
+叩
+拽
+幅
+硬
+惶
+桀
+漠
+措
+泼
+唑
+齐
+肾
+念
+酱
+虚
+屁
+耶
+旗
+砦
+闵
+婉
+馆
+拭
+绅
+韧
+忏
+窝
+醋
+葺
+顾
+辞
+倜
+堆
+辋
+逆
+玟
+贱
+疾
+董
+惘
+倌
+锕
+淘
+嘀
+莽
+俭
+笏
+绑
+鲷
+杈
+择
+蟀
+粥
+嗯
+驰
+逾
+案
+谪
+褓
+胫
+哩
+昕
+颚
+鲢
+绠
+躺
+鹄
+崂
+儒
+俨
+丝
+尕
+泌
+啊
+萸
+彰
+幺
+吟
+骄
+苣
+弦
+脊
+瑰
+〈
+诛
+镁
+析
+闪
+剪
+侧
+哟
+框
+螃
+守
+嬗
+燕
+狭
+铈
+缮
+概
+迳
+痧
+鲲
+俯
+售
+笼
+痣
+扉
+挖
+满
+咋
+援
+邱
+扇
+歪
+便
+玑
+绦
+峡
+蛇
+叨
+〖
+泽
+胃
+斓
+喋
+怂
+坟
+猪
+该
+蚬
+炕
+弥
+赞
+棣
+晔
+娠
+挲
+狡
+创
+疖
+铕
+镭
+稷
+挫
+弭
+啾
+翔
+粉
+履
+苘
+哦
+楼
+秕
+铂
+土
+锣
+瘟
+挣
+栉
+习
+享
+桢
+袅
+磨
+桂
+谦
+延
+坚
+蔚
+噗
+署
+谟
+猬
+钎
+恐
+嬉
+雒
+倦
+衅
+亏
+璩
+睹
+刻
+殿
+王
+算
+雕
+麻
+丘
+柯
+骆
+丸
+塍
+谚
+添
+鲈
+垓
+桎
+蚯
+芥
+予
+飕
+镦
+谌
+窗
+醚
+菀
+亮
+搪
+莺
+蒿
+羁
+足
+J
+真
+轶
+悬
+衷
+靛
+翊
+掩
+哒
+炅
+掐
+冼
+妮
+l
+谐
+稚
+荆
+擒
+犯
+陵
+虏
+浓
+崽
+刍
+陌
+傻
+孜
+千
+靖
+演
+矜
+钕
+煽
+杰
+酗
+渗
+伞
+栋
+俗
+泫
+戍
+罕
+沾
+疽
+灏
+煦
+芬
+磴
+叱
+阱
+榉
+湃
+蜀
+叉
+醒
+彪
+租
+郡
+篷
+屎
+良
+垢
+隗
+弱
+陨
+峪
+砷
+掴
+颁
+胎
+雯
+绵
+贬
+沐
+撵
+隘
+篙
+暖
+曹
+陡
+栓
+填
+臼
+彦
+瓶
+琪
+潼
+哪
+鸡
+摩
+啦
+俟
+锋
+域
+耻
+蔫
+疯
+纹
+撇
+毒
+绶
+痛
+酯
+忍
+爪
+赳
+歆
+嘹
+辕
+烈
+册
+朴
+钱
+吮
+毯
+癜
+娃
+谀
+邵
+厮
+炽
+璞
+邃
+丐
+追
+词
+瓒
+忆
+轧
+芫
+谯
+喷
+弟
+半
+冕
+裙
+掖
+墉
+绮
+寝
+苔
+势
+顷
+褥
+切
+衮
+君
+佳
+嫒
+蚩
+霞
+佚
+洙
+逊
+镖
+暹
+唛
+&
+殒
+顶
+碗
+獗
+轭
+铺
+蛊
+废
+恹
+汨
+崩
+珍
+那
+杵
+曲
+纺
+夏
+薰
+傀
+闳
+淬
+姘
+舀
+拧
+卷
+楂
+恍
+讪
+厩
+寮
+篪
+赓
+乘
+灭
+盅
+鞣
+沟
+慎
+挂
+饺
+鼾
+杳
+树
+缨
+丛
+絮
+娌
+臻
+嗳
+篡
+侩
+述
+衰
+矛
+圈
+蚜
+匕
+筹
+匿
+濞
+晨
+叶
+骋
+郝
+挚
+蚴
+滞
+增
+侍
+描
+瓣
+吖
+嫦
+蟒
+匾
+圣
+赌
+毡
+癞
+恺
+百
+曳
+需
+篓
+肮
+庖
+帏
+卿
+驿
+遗
+蹬
+鬓
+骡
+歉
+芎
+胳
+屐
+禽
+烦
+晌
+寄
+媾
+狄
+翡
+苒
+船
+廉
+终
+痞
+殇
+々
+畦
+饶
+改
+拆
+悻
+萄
+￡
+瓿
+乃
+訾
+桅
+匮
+溧
+拥
+纱
+铍
+骗
+蕃
+龋
+缬
+父
+佐
+疚
+栎
+醍
+掳
+蓄
+x
+惆
+颜
+鲆
+榆
+〔
+猎
+敌
+暴
+谥
+鲫
+贾
+罗
+玻
+缄
+扦
+芪
+癣
+落
+徒
+臾
+恿
+猩
+托
+邴
+肄
+牵
+春
+陛
+耀
+刊
+拓
+蓓
+邳
+堕
+寇
+枉
+淌
+啡
+湄
+兽
+酷
+萼
+碚
+濠
+萤
+夹
+旬
+戮
+梭
+琥
+椭
+昔
+勺
+蜊
+绐
+晚
+孺
+僵
+宣
+摄
+冽
+旨
+萌
+忙
+蚤
+眉
+噼
+蟑
+付
+契
+瓜
+悼
+颡
+壁
+曾
+窕
+颢
+澎
+仿
+俑
+浑
+嵌
+浣
+乍
+碌
+褪
+乱
+蔟
+隙
+玩
+剐
+葫
+箫
+纲
+围
+伐
+决
+伙
+漩
+瑟
+刑
+肓
+镳
+缓
+蹭
+氨
+皓
+典
+畲
+坍
+铑
+檐
+塑
+洞
+倬
+储
+胴
+淳
+戾
+吐
+灼
+惺
+妙
+毕
+珐
+缈
+虱
+盖
+羰
+鸿
+磅
+谓
+髅
+娴
+苴
+唷
+蚣
+霹
+抨
+贤
+唠
+犬
+誓
+逍
+庠
+逼
+麓
+籼
+釉
+呜
+碧
+秧
+氩
+摔
+霄
+穸
+纨
+辟
+妈
+映
+完
+牛
+缴
+嗷
+炊
+恩
+荔
+茆
+掉
+紊
+慌
+莓
+羟
+阙
+萁
+磐
+另
+蕹
+辱
+鳐
+湮
+吡
+吩
+唐
+睦
+垠
+舒
+圜
+冗
+瞿
+溺
+芾
+囱
+匠
+僳
+汐
+菩
+饬
+漓
+黑
+霰
+浸
+濡
+窥
+毂
+蒡
+兢
+驻
+鹉
+芮
+诙
+迫
+雳
+厂
+忐
+臆
+猴
+鸣
+蚪
+栈
+箕
+羡
+渐
+莆
+捍
+眈
+哓
+趴
+蹼
+埕
+嚣
+骛
+宏
+淄
+斑
+噜
+严
+瑛
+垃
+椎
+诱
+压
+庾
+绞
+焘
+廿
+抡
+迄
+棘
+夫
+纬
+锹
+眨
+瞌
+侠
+脐
+竞
+瀑
+孳
+骧
+遁
+姜
+颦
+荪
+滚
+萦
+伪
+逸
+粳
+爬
+锁
+矣
+役
+趣
+洒
+颔
+诏
+逐
+奸
+甭
+惠
+攀
+蹄
+泛
+尼
+拼
+阮
+鹰
+亚
+颈
+惑
+勒
+〉
+际
+肛
+爷
+刚
+钨
+丰
+养
+冶
+鲽
+辉
+蔻
+画
+覆
+皴
+妊
+麦
+返
+醉
+皂
+擀
+〗
+酶
+凑
+粹
+悟
+诀
+硖
+港
+卜
+z
+杀
+涕
+±
+舍
+铠
+抵
+弛
+段
+敝
+镐
+奠
+拂
+轴
+跛
+袱
+e
+t
+沉
+菇
+俎
+薪
+峦
+秭
+蟹
+历
+盟
+菠
+寡
+液
+肢
+喻
+染
+裱
+悱
+抱
+氙
+赤
+捅
+猛
+跑
+氮
+谣
+仁
+尺
+辊
+窍
+烙
+衍
+架
+擦
+倏
+璐
+瑁
+币
+楞
+胖
+夔
+趸
+邛
+惴
+饕
+虔
+蝎
+§
+哉
+贝
+宽
+辫
+炮
+扩
+饲
+籽
+魏
+菟
+锰
+伍
+猝
+末
+琳
+哚
+蛎
+邂
+呀
+姿
+鄞
+却
+歧
+仙
+恸
+椐
+森
+牒
+寤
+袒
+婆
+虢
+雅
+钉
+朵
+贼
+欲
+苞
+寰
+故
+龚
+坭
+嘘
+咫
+礼
+硷
+兀
+睢
+汶
+’
+铲
+烧
+绕
+诃
+浃
+钿
+哺
+柜
+讼
+颊
+璁
+腔
+洽
+咐
+脲
+簌
+筠
+镣
+玮
+鞠
+谁
+兼
+姆
+挥
+梯
+蝴
+谘
+漕
+刷
+躏
+宦
+弼
+b
+垌
+劈
+麟
+莉
+揭
+笙
+渎
+仕
+嗤
+仓
+配
+怏
+抬
+错
+泯
+镊
+孰
+猿
+邪
+仍
+秋
+鼬
+壹
+歇
+吵
+炼
+<
+尧
+射
+柬
+廷
+胧
+霾
+凳
+隋
+肚
+浮
+梦
+祥
+株
+堵
+退
+L
+鹫
+跎
+凶
+毽
+荟
+炫
+栩
+玳
+甜
+沂
+鹿
+顽
+伯
+爹
+赔
+蛴
+徐
+匡
+欣
+狰
+缸
+雹
+蟆
+疤
+默
+沤
+啜
+痂
+衣
+禅
+w
+i
+h
+辽
+葳
+黝
+钗
+停
+沽
+棒
+馨
+颌
+肉
+吴
+硫
+悯
+劾
+娈
+马
+啧
+吊
+悌
+镑
+峭
+帆
+瀣
+涉
+咸
+疸
+滋
+泣
+翦
+拙
+癸
+钥
+蜒
++
+尾
+庄
+凝
+泉
+婢
+渴
+谊
+乞
+陆
+锉
+糊
+鸦
+淮
+I
+B
+N
+晦
+弗
+乔
+庥
+葡
+尻
+席
+橡
+傣
+渣
+拿
+惩
+麋
+斛
+缃
+矮
+蛏
+岘
+鸽
+姐
+膏
+催
+奔
+镒
+喱
+蠡
+摧
+钯
+胤
+柠
+拐
+璋
+鸥
+卢
+荡
+倾
+^
+_
+珀
+逄
+萧
+塾
+掇
+贮
+笆
+聂
+圃
+冲
+嵬
+M
+滔
+笕
+值
+炙
+偶
+蜱
+搐
+梆
+汪
+蔬
+腑
+鸯
+蹇
+敞
+绯
+仨
+祯
+谆
+梧
+糗
+鑫
+啸
+豺
+囹
+猾
+巢
+柄
+瀛
+筑
+踌
+沭
+暗
+苁
+鱿
+蹉
+脂
+蘖
+牢
+热
+木
+吸
+溃
+宠
+序
+泞
+偿
+拜
+檩
+厚
+朐
+毗
+螳
+吞
+媚
+朽
+担
+蝗
+橘
+畴
+祈
+糟
+盱
+隼
+郜
+惜
+珠
+裨
+铵
+焙
+琚
+唯
+咚
+噪
+骊
+丫
+滢
+勤
+棉
+呸
+咣
+淀
+隔
+蕾
+窈
+饨
+挨
+煅
+短
+匙
+粕
+镜
+赣
+撕
+墩
+酬
+馁
+豌
+颐
+抗
+酣
+氓
+佑
+搁
+哭
+递
+耷
+涡
+桃
+贻
+碣
+截
+瘦
+昭
+镌
+蔓
+氚
+甲
+猕
+蕴
+蓬
+散
+拾
+纛
+狼
+猷
+铎
+埋
+旖
+矾
+讳
+囊
+糜
+迈
+粟
+蚂
+紧
+鲳
+瘢
+栽
+稼
+羊
+锄
+斟
+睁
+桥
+瓮
+蹙
+祉
+醺
+鼻
+昱
+剃
+跳
+篱
+跷
+蒜
+翎
+宅
+晖
+嗑
+壑
+峻
+癫
+屏
+狠
+陋
+袜
+途
+憎
+祀
+莹
+滟
+佶
+溥
+臣
+约
+盛
+峰
+磁
+慵
+婪
+拦
+莅
+朕
+鹦
+粲
+裤
+哎
+疡
+嫖
+琵
+窟
+堪
+谛
+嘉
+儡
+鳝
+斩
+郾
+驸
+酊
+妄
+胜
+贺
+徙
+傅
+噌
+钢
+栅
+庇
+恋
+匝
+巯
+邈
+尸
+锚
+粗
+佟
+蛟
+薹
+纵
+蚊
+郅
+绢
+锐
+苗
+俞
+篆
+淆
+膀
+鲜
+煎
+诶
+秽
+寻
+涮
+刺
+怀
+噶
+巨
+褰
+魅
+灶
+灌
+桉
+藕
+谜
+舸
+薄
+搀
+恽
+借
+牯
+痉
+渥
+愿
+亓
+耘
+杠
+柩
+锔
+蚶
+钣
+珈
+喘
+蹒
+幽
+赐
+稗
+晤
+莱
+泔
+扯
+肯
+菪
+裆
+腩
+豉
+疆
+骜
+腐
+倭
+珏
+唔
+粮
+亡
+润
+慰
+伽
+橄
+玄
+誉
+醐
+胆
+龊
+粼
+塬
+陇
+彼
+削
+嗣
+绾
+芽
+妗
+垭
+瘴
+爽
+薏
+寨
+龈
+泠
+弹
+赢
+漪
+猫
+嘧
+涂
+恤
+圭
+茧
+烽
+屑
+痕
+巾
+赖
+荸
+凰
+腮
+畈
+亵
+蹲
+偃
+苇
+澜
+艮
+换
+骺
+烘
+苕
+梓
+颉
+肇
+哗
+悄
+氤
+涠
+葬
+屠
+鹭
+植
+竺
+佯
+诣
+鲇
+瘀
+鲅
+邦
+移
+滁
+冯
+耕
+癔
+戌
+茬
+沁
+巩
+悠
+湘
+洪
+痹
+锟
+循
+谋
+腕
+鳃
+钠
+捞
+焉
+迎
+碱
+伫
+急
+榷
+奈
+邝
+卯
+辄
+皲
+卟
+醛
+畹
+忧
+稳
+雄
+昼
+缩
+阈
+睑
+扌
+耗
+曦
+涅
+捏
+瞧
+邕
+淖
+漉
+铝
+耦
+禹
+湛
+喽
+莼
+琅
+诸
+苎
+纂
+硅
+始
+嗨
+傥
+燃
+臂
+赅
+嘈
+呆
+贵
+屹
+壮
+肋
+亍
+蚀
+卅
+豹
+腆
+邬
+迭
+浊
+}
+童
+螂
+捐
+圩
+勐
+触
+寞
+汊
+壤
+荫
+膺
+渌
+芳
+懿
+遴
+螈
+泰
+蓼
+蛤
+茜
+舅
+枫
+朔
+膝
+眙
+避
+梅
+判
+鹜
+璜
+牍
+缅
+垫
+藻
+黔
+侥
+惚
+懂
+踩
+腰
+腈
+札
+丞
+唾
+慈
+顿
+摹
+荻
+琬
+~
+斧
+沈
+滂
+胁
+胀
+幄
+莜
+Z
+匀
+鄄
+掌
+绰
+茎
+焚
+赋
+萱
+谑
+汁
+铒
+瞎
+夺
+蜗
+野
+娆
+冀
+弯
+篁
+懵
+灞
+隽
+芡
+脘
+俐
+辩
+芯
+掺
+喏
+膈
+蝈
+觐
+悚
+踹
+蔗
+熠
+鼠
+呵
+抓
+橼
+峨
+畜
+缔
+禾
+崭
+弃
+熊
+摒
+凸
+拗
+穹
+蒙
+抒
+祛
+劝
+闫
+扳
+阵
+醌
+踪
+喵
+侣
+搬
+仅
+荧
+赎
+蝾
+琦
+买
+婧
+瞄
+寓
+皎
+冻
+赝
+箩
+莫
+瞰
+郊
+笫
+姝
+筒
+枪
+遣
+煸
+袋
+舆
+痱
+涛
+母
+〇
+启
+践
+耙
+绲
+盘
+遂
+昊
+搞
+槿
+诬
+纰
+泓
+惨
+檬
+亻
+越
+C
+o
+憩
+熵
+祷
+钒
+暧
+塔
+阗
+胰
+咄
+娶
+魔
+琶
+钞
+邻
+扬
+杉
+殴
+咽
+弓
+〆
+髻
+】
+吭
+揽
+霆
+拄
+殖
+脆
+彻
+岩
+芝
+勃
+辣
+剌
+钝
+嘎
+甄
+佘
+皖
+伦
+授
+徕
+憔
+挪
+皇
+庞
+稔
+芜
+踏
+溴
+兖
+卒
+擢
+饥
+鳞
+煲
+‰
+账
+颗
+叻
+斯
+捧
+鳍
+琮
+讹
+蛙
+纽
+谭
+酸
+兔
+莒
+睇
+伟
+觑
+羲
+嗜
+宜
+褐
+旎
+辛
+卦
+诘
+筋
+鎏
+溪
+挛
+熔
+阜
+晰
+鳅
+丢
+奚
+灸
+呱
+献
+陉
+黛
+鸪
+甾
+萨
+疮
+拯
+洲
+疹
+辑
+叙
+恻
+谒
+允
+柔
+烂
+氏
+逅
+漆
+拎
+惋
+扈
+湟
+纭
+啕
+掬
+擞
+哥
+忽
+涤
+鸵
+靡
+郗
+瓷
+扁
+廊
+怨
+雏
+钮
+敦
+E
+懦
+憋
+汀
+拚
+啉
+腌
+岸
+f
+痼
+瞅
+尊
+咀
+眩
+飙
+忌
+仝
+迦
+熬
+毫
+胯
+篑
+茄
+腺
+凄
+舛
+碴
+锵
+诧
+羯
+後
+漏
+汤
+宓
+仞
+蚁
+壶
+谰
+皑
+铄
+棰
+罔
+辅
+晶
+苦
+牟
+闽
+\
+烃
+饮
+聿
+丙
+蛳
+朱
+煤
+涔
+鳖
+犁
+罐
+荼
+砒
+淦
+妤
+黏
+戎
+孑
+婕
+瑾
+戢
+钵
+枣
+捋
+砥
+衩
+狙
+桠
+稣
+阎
+肃
+梏
+诫
+孪
+昶
+婊
+衫
+嗔
+侃
+塞
+蜃
+樵
+峒
+貌
+屿
+欺
+缫
+阐
+栖
+诟
+珞
+荭
+吝
+萍
+嗽
+恂
+啻
+蜴
+磬
+峋
+俸
+豫
+谎
+徊
+镍
+韬
+魇
+晴
+U
+囟
+猜
+蛮
+坐
+囿
+伴
+亭
+肝
+佗
+蝠
+妃
+胞
+滩
+榴
+氖
+垩
+苋
+砣
+扪
+馏
+姓
+轩
+厉
+夥
+侈
+禀
+垒
+岑
+赏
+钛
+辐
+痔
+披
+纸
+碳
+“
+坞
+蠓
+挤
+荥
+沅
+悔
+铧
+帼
+蒌
+蝇
+a
+p
+y
+n
+g
+哀
+浆
+瑶
+凿
+桶
+馈
+皮
+奴
+苜
+佤
+伶
+晗
+铱
+炬
+优
+弊
+氢
+恃
+甫
+攥
+端
+锌
+灰
+稹
+炝
+曙
+邋
+亥
+眶
+碾
+拉
+萝
+绔
+捷
+浍
+腋
+姑
+菖
+凌
+涞
+麽
+锢
+桨
+潢
+绎
+镰
+殆
+锑
+渝
+铬
+困
+绽
+觎
+匈
+糙
+暑
+裹
+鸟
+盔
+肽
+迷
+綦
+『
+亳
+佝
+俘
+钴
+觇
+骥
+仆
+疝
+跪
+婶
+郯
+瀹
+唉
+脖
+踞
+针
+晾
+忒
+扼
+瞩
+叛
+椒
+疟
+嗡
+邗
+肆
+跆
+玫
+忡
+捣
+咧
+唆
+艄
+蘑
+潦
+笛
+阚
+沸
+泻
+掊
+菽
+贫
+斥
+髂
+孢
+镂
+赂
+麝
+鸾
+屡
+衬
+苷
+恪
+叠
+希
+粤
+爻
+喝
+茫
+惬
+郸
+绻
+庸
+撅
+碟
+宄
+妹
+膛
+叮
+饵
+崛
+嗲
+椅
+冤
+搅
+咕
+敛
+尹
+垦
+闷
+蝉
+霎
+勰
+败
+蓑
+泸
+肤
+鹌
+幌
+焦
+浠
+鞍
+刁
+舰
+乙
+竿
+裔
+。
+茵
+函
+伊
+兄
+丨
+娜
+匍
+謇
+莪
+宥
+似
+蝽
+翳
+酪
+翠
+粑
+薇
+祢
+骏
+赠
+叫
+Q
+噤
+噻
+竖
+芗
+莠
+潭
+俊
+羿
+耜
+O
+郫
+趁
+嗪
+囚
+蹶
+芒
+洁
+笋
+鹑
+敲
+硝
+啶
+堡
+渲
+揩
+』
+携
+宿
+遒
+颍
+扭
+棱
+割
+萜
+蔸
+葵
+琴
+捂
+饰
+衙
+耿
+掠
+募
+岂
+窖
+涟
+蔺
+瘤
+柞
+瞪
+怜
+匹
+距
+楔
+炜
+哆
+秦
+缎
+幼
+茁
+绪
+痨
+恨
+楸
+娅
+瓦
+桩
+雪
+嬴
+伏
+榔
+妥
+铿
+拌
+眠
+雍
+缇
+‘
+卓
+搓
+哌
+觞
+噩
+屈
+哧
+髓
+咦
+巅
+娑
+侑
+淫
+膳
+祝
+勾
+姊
+莴
+胄
+疃
+薛
+蜷
+胛
+巷
+芙
+芋
+熙
+闰
+勿
+窃
+狱
+剩
+钏
+幢
+陟
+铛
+慧
+靴
+耍
+k
+浙
+浇
+飨
+惟
+绗
+祜
+澈
+啼
+咪
+磷
+摞
+诅
+郦
+抹
+跃
+壬
+吕
+肖
+琏
+颤
+尴
+剡
+抠
+凋
+赚
+泊
+津
+宕
+殷
+倔
+氲
+漫
+邺
+涎
+怠
+$
+垮
+荬
+遵
+俏
+叹
+噢
+饽
+蜘
+孙
+筵
+疼
+鞭
+羧
+牦
+箭
+潴
+c
+眸
+祭
+髯
+啖
+坳
+愁
+芩
+驮
+倡
+巽
+穰
+沃
+胚
+怒
+凤
+槛
+剂
+趵
+嫁
+v
+邢
+灯
+鄢
+桐
+睽
+檗
+锯
+槟
+婷
+嵋
+圻
+诗
+蕈
+颠
+遭
+痢
+芸
+怯
+馥
+竭
+锗
+徜
+恭
+遍
+籁
+剑
+嘱
+苡
+龄
+僧
+桑
+潸
+弘
+澶
+楹
+悲
+讫
+愤
+腥
+悸
+谍
+椹
+呢
+桓
+葭
+攫
+阀
+翰
+躲
+敖
+柑
+郎
+笨
+橇
+呃
+魁
+燎
+脓
+葩
+磋
+垛
+玺
+狮
+沓
+砜
+蕊
+锺
+罹
+蕉
+翱
+虐
+闾
+巫
+旦
+茱
+嬷
+枯
+鹏
+贡
+芹
+汛
+矫
+绁
+拣
+禺
+佃
+讣
+舫
+惯
+乳
+趋
+疲
+挽
+岚
+虾
+衾
+蠹
+蹂
+飓
+氦
+铖
+孩
+稞
+瑜
+壅
+掀
+勘
+妓
+畅
+髋
+W
+庐
+牲
+蓿
+榕
+练
+垣
+唱
+邸
+菲
+昆
+婺
+穿
+绡
+麒
+蚱
+掂
+愚
+泷
+涪
+漳
+妩
+娉
+榄
+讷
+觅
+旧
+藤
+煮
+呛
+柳
+腓
+叭
+庵
+烷
+阡
+罂
+蜕
+擂
+猖
+咿
+媲
+脉
+【
+沏
+貅
+黠
+熏
+哲
+烁
+坦
+酵
+兜
+×
+潇
+撒
+剽
+珩
+圹
+乾
+摸
+樟
+帽
+嗒
+襄
+魂
+轿
+憬
+锡
+〕
+喃
+皆
+咖
+隅
+脸
+残
+泮
+袂
+鹂
+珊
+囤
+捆
+咤
+误
+徨
+闹
+淙
+芊
+淋
+怆
+囗
+拨
+梳
+渤
+R
+G
+绨
+蚓
+婀
+幡
+狩
+麾
+谢
+唢
+裸
+旌
+伉
+纶
+裂
+驳
+砼
+咛
+澄
+樨
+蹈
+宙
+澍
+倍
+貔
+操
+勇
+蟠
+摈
+砧
+虬
+够
+缁
+悦
+藿
+撸
+艹
+摁
+淹
+豇
+虎
+榭
+ˉ
+吱
+d
+°
+喧
+荀
+踱
+侮
+奋
+偕
+饷
+犍
+惮
+坑
+璎
+徘
+宛
+妆
+袈
+倩
+窦
+昂
+荏
+乖
+K
+怅
+撰
+鳙
+牙
+袁
+酞
+X
+痿
+琼
+闸
+雁
+趾
+荚
+虻
+涝
+《
+杏
+韭
+偈
+烤
+绫
+鞘
+卉
+症
+遢
+蓥
+诋
+杭
+荨
+匆
+竣
+簪
+辙
+敕
+虞
+丹
+缭
+咩
+黟
+m
+淤
+瑕
+咂
+铉
+硼
+茨
+嶂
+痒
+畸
+敬
+涿
+粪
+窘
+熟
+叔
+嫔
+盾
+忱
+裘
+憾
+梵
+赡
+珙
+咯
+娘
+庙
+溯
+胺
+葱
+痪
+摊
+荷
+卞
+乒
+髦
+寐
+铭
+坩
+胗
+枷
+爆
+溟
+嚼
+羚
+砬
+轨
+惊
+挠
+罄
+竽
+菏
+氧
+浅
+楣
+盼
+枢
+炸
+阆
+杯
+谏
+噬
+淇
+渺
+俪
+秆
+墓
+泪
+跻
+砌
+痰
+垡
+渡
+耽
+釜
+讶
+鳎
+煞
+呗
+韶
+舶
+绷
+鹳
+缜
+旷
+铊
+皱
+龌
+檀
+霖
+奄
+槐
+艳
+蝶
+旋
+哝
+赶
+骞
+蚧
+腊
+盈
+丁
+`
+蜚
+矸
+蝙
+睨
+嚓
+僻
+鬼
+醴
+夜
+彝
+磊
+笔
+拔
+栀
+糕
+厦
+邰
+纫
+逭
+纤
+眦
+膊
+馍
+躇
+烯
+蘼
+冬
+诤
+暄
+骶
+哑
+瘠
+」
+臊
+丕
+愈
+咱
+螺
+擅
+跋
+搏
+硪
+谄
+笠
+淡
+嘿
+骅
+谧
+鼎
+皋
+姚
+歼
+蠢
+驼
+耳
+胬
+挝
+涯
+狗
+蒽
+孓
+犷
+凉
+芦
+箴
+铤
+孤
+嘛
+坤
+V
+茴
+朦
+挞
+尖
+橙
+诞
+搴
+碇
+洵
+浚
+帚
+蜍
+漯
+柘
+嚎
+讽
+芭
+荤
+咻
+祠
+秉
+跖
+埃
+吓
+糯
+眷
+馒
+惹
+娼
+鲑
+嫩
+讴
+轮
+瞥
+靶
+褚
+乏
+缤
+宋
+帧
+删
+驱
+碎
+扑
+俩
+俄
+偏
+涣
+竹
+噱
+皙
+佰
+渚
+唧
+斡
+#
+镉
+刀
+崎
+筐
+佣
+夭
+贰
+肴
+峙
+哔
+艿
+匐
+牺
+镛
+缘
+仡
+嫡
+劣
+枸
+堀
+梨
+簿
+鸭
+蒸
+亦
+稽
+浴
+{
+衢
+束
+槲
+j
+阁
+揍
+疥
+棋
+潋
+聪
+窜
+乓
+睛
+插
+冉
+阪
+苍
+搽
+「
+蟾
+螟
+幸
+仇
+樽
+撂
+慢
+跤
+幔
+俚
+淅
+覃
+觊
+溶
+妖
+帛
+侨
+曰
+妾
+泗
+·
+：
+瀘
+風
+Ë
+（
+）
+∶
+紅
+紗
+瑭
+雲
+頭
+鶏
+財
+許
+•
+¥
+樂
+焗
+麗
+—
+；
+滙
+東
+榮
+繪
+興
+…
+門
+業
+π
+楊
+國
+顧
+é
+盤
+寳
+Λ
+龍
+鳳
+島
+誌
+緣
+結
+銭
+萬
+勝
+祎
+璟
+優
+歡
+臨
+時
+購
+＝
+★
+藍
+昇
+鐵
+觀
+勅
+農
+聲
+畫
+兿
+術
+發
+劉
+記
+專
+耑
+園
+書
+壴
+種
+Ο
+●
+褀
+號
+銀
+匯
+敟
+锘
+葉
+橪
+廣
+進
+蒄
+鑽
+阝
+祙
+貢
+鍋
+豊
+夬
+喆
+團
+閣
+開
+燁
+賓
+館
+酡
+沔
+順
+＋
+硚
+劵
+饸
+陽
+車
+湓
+復
+萊
+氣
+軒
+華
+堃
+迮
+纟
+戶
+馬
+學
+裡
+電
+嶽
+獨
+マ
+シ
+サ
+ジ
+燘
+袪
+環
+❤
+臺
+灣
+専
+賣
+孖
+聖
+攝
+線
+▪
+α
+傢
+俬
+夢
+達
+莊
+喬
+貝
+薩
+劍
+羅
+壓
+棛
+饦
+尃
+璈
+囍
+醫
+Ｇ
+Ｉ
+Ａ
+＃
+Ｎ
+鷄
+髙
+嬰
+啓
+約
+隹
+潔
+賴
+藝
+～
+寶
+籣
+麺
+　
+嶺
+√
+義
+網
+峩
+長
+∧
+魚
+機
+構
+②
+鳯
+偉
+Ｌ
+Ｂ
+㙟
+畵
+鴿
+＇
+詩
+溝
+嚞
+屌
+藔
+佧
+玥
+蘭
+織
+１
+３
+９
+０
+７
+點
+砭
+鴨
+鋪
+銘
+廳
+弍
+‧
+創
+湯
+坶
+℃
+卩
+骝
+＆
+烜
+荘
+當
+潤
+扞
+係
+懷
+碶
+钅
+蚨
+讠
+☆
+叢
+爲
+埗
+涫
+塗
+→
+楽
+現
+鯨
+愛
+瑪
+鈺
+忄
+悶
+藥
+飾
+樓
+視
+孬
+ㆍ
+燚
+苪
+師
+①
+丼
+锽
+│
+韓
+標
+è
+兒
+閏
+匋
+張
+漢
+Ü
+髪
+會
+閑
+檔
+習
+裝
+の
+峯
+菘
+輝
+И
+雞
+釣
+億
+浐
+Ｋ
+Ｏ
+Ｒ
+８
+Ｈ
+Ｅ
+Ｐ
+Ｔ
+Ｗ
+Ｄ
+Ｓ
+Ｃ
+Ｍ
+Ｆ
+姌
+饹
+»
+晞
+廰
+ä
+嵯
+鷹
+負
+飲
+絲
+冚
+楗
+澤
+綫
+區
+❋
+←
+質
+靑
+揚
+③
+滬
+統
+産
+協
+﹑
+乸
+畐
+經
+運
+際
+洺
+岽
+為
+粵
+諾
+崋
+豐
+碁
+ɔ
+Ｖ
+２
+６
+齋
+誠
+訂
+´
+勑
+雙
+陳
+無
+í
+泩
+媄
+夌
+刂
+ｉ
+ｃ
+ｔ
+ｏ
+ｒ
+ａ
+嘢
+耄
+燴
+暃
+壽
+媽
+靈
+抻
+體
+唻
+É
+冮
+甹
+鎮
+錦
+ʌ
+蜛
+蠄
+尓
+駕
+戀
+飬
+逹
+倫
+貴
+極
+Я
+Й
+寬
+磚
+嶪
+郎
+職
+｜
+間
+ｎ
+ｄ
+剎
+伈
+課
+飛
+橋
+瘊
+№
+譜
+骓
+圗
+滘
+縣
+粿
+咅
+養
+濤
+彳
+®
+％
+Ⅱ
+啰
+㴪
+見
+矞
+薬
+糁
+邨
+鲮
+顔
+罱
+З
+選
+話
+贏
+氪
+俵
+競
+瑩
+繡
+枱
+β
+綉
+á
+獅
+爾
+™
+麵
+戋
+淩
+徳
+個
+劇
+場
+務
+簡
+寵
+ｈ
+實
+膠
+轱
+圖
+築
+嘣
+樹
+㸃
+營
+耵
+孫
+饃
+鄺
+飯
+麯
+遠
+輸
+坫
+孃
+乚
+閃
+鏢
+㎡
+題
+廠
+關
+↑
+爺
+將
+軍
+連
+篦
+覌
+參
+箸
+－
+窠
+棽
+寕
+夀
+爰
+歐
+呙
+閥
+頡
+熱
+雎
+垟
+裟
+凬
+勁
+帑
+馕
+夆
+疌
+枼
+馮
+貨
+蒤
+樸
+彧
+旸
+靜
+龢
+暢
+㐱
+鳥
+珺
+鏡
+灡
+爭
+堷
+廚
+Ó
+騰
+診
+┅
+蘇
+褔
+凱
+頂
+豕
+亞
+帥
+嘬
+⊥
+仺
+桖
+複
+饣
+絡
+穂
+顏
+棟
+納
+▏
+濟
+親
+設
+計
+攵
+埌
+烺
+ò
+頤
+燦
+蓮
+撻
+節
+講
+濱
+濃
+娽
+洳
+朿
+燈
+鈴
+護
+膚
+铔
+過
+補
+Ｚ
+Ｕ
+５
+４
+坋
+闿
+䖝
+餘
+缐
+铞
+貿
+铪
+桼
+趙
+鍊
+［
+㐂
+垚
+菓
+揸
+捲
+鐘
+滏
+𣇉
+爍
+輪
+燜
+鴻
+鮮
+動
+鹞
+鷗
+丄
+慶
+鉌
+翥
+飮
+腸
+⇋
+漁
+覺
+來
+熘
+昴
+翏
+鲱
+圧
+鄉
+萭
+頔
+爐
+嫚
+г
+貭
+類
+聯
+幛
+輕
+訓
+鑒
+夋
+锨
+芃
+珣
+䝉
+扙
+嵐
+銷
+處
+ㄱ
+���
+誘
+苝
+歸
+儀
+燒
+楿
+內
+粢
+葒
+奧
+麥
+礻
+滿
+蠔
+穵
+瞭
+態
+鱬
+榞
+硂
+鄭
+黃
+煙
+祐
+奓
+逺
+＊
+瑄
+獲
+聞
+薦
+讀
+這
+樣
+決
+問
+啟
+們
+執
+説
+轉
+單
+隨
+唘
+帶
+倉
+庫
+還
+贈
+尙
+皺
+■
+餅
+產
+○
+∈
+報
+狀
+楓
+賠
+琯
+嗮
+禮
+｀
+傳
+＞
+≤
+嗞
+Φ
+≥
+換
+咭
+∣
+↓
+曬
+ε
+応
+寫
+″
+終
+様
+純
+費
+療
+聨
+凍
+壐
+郵
+ü
+黒
+∫
+製
+塊
+調
+軽
+確
+撃
+級
+馴
+Ⅲ
+涇
+繹
+數
+碼
+證
+狒
+処
+劑
+＜
+晧
+賀
+衆
+］
+櫥
+兩
+陰
+絶
+對
+鯉
+憶
+◎
+ｐ
+ｅ
+Ｙ
+蕒
+煖
+頓
+測
+試
+鼽
+僑
+碩
+妝
+帯
+≈
+鐡
+舖
+權
+喫
+倆
+ˋ
+該
+悅
+ā
+俫
+．
+ｆ
+ｓ
+ｂ
+ｍ
+ｋ
+ｇ
+ｕ
+ｊ
+貼
+淨
+濕
+針
+適
+備
+ｌ
+／
+給
+謢
+強
+觸
+衛
+與
+⊙
+＄
+緯
+變
+⑴
+⑵
+⑶
+㎏
+殺
+∩
+幚
+─
+價
+▲
+離
+ú
+ó
+飄
+烏
+関
+閟
+﹝
+﹞
+邏
+輯
+鍵
+驗
+訣
+導
+歷
+屆
+層
+▼
+儱
+錄
+熳
+ē
+艦
+吋
+錶
+辧
+飼
+顯
+④
+禦
+販
+気
+対
+枰
+閩
+紀
+幹
+瞓
+貊
+淚
+△
+眞
+墊
+Ω
+獻
+褲
+縫
+緑
+亜
+鉅
+餠
+｛
+｝
+◆
+蘆
+薈
+█
+◇
+溫
+彈
+晳
+粧
+犸
+穩
+訊
+崬
+凖
+熥
+П
+舊
+條
+紋
+圍
+Ⅳ
+筆
+尷
+難
+雜
+錯
+綁
+識
+頰
+鎖
+艶
+□
+殁
+殼
+⑧
+├
+▕
+鵬
+ǐ
+ō
+ǒ
+糝
+綱
+▎
+μ
+盜
+饅
+醬
+籤
+蓋
+釀
+鹽
+據
+à
+ɡ
+辦
+◥
+彐
+┌
+婦
+獸
+鲩
+伱
+ī
+蒟
+蒻
+齊
+袆
+腦
+寧
+凈
+妳
+煥
+詢
+偽
+謹
+啫
+鯽
+騷
+鱸
+損
+傷
+鎻
+髮
+買
+冏
+儥
+両
+﹢
+∞
+載
+喰
+ｚ
+羙
+悵
+燙
+曉
+員
+組
+徹
+艷
+痠
+鋼
+鼙
+縮
+細
+嚒
+爯
+≠
+維
+＂
+鱻
+壇
+厍
+帰
+浥
+犇
+薡
+軎
+²
+應
+醜
+刪
+緻
+鶴
+賜
+噁
+軌
+尨
+镔
+鷺
+槗
+彌
+葚
+濛
+請
+溇
+緹
+賢
+訪
+獴
+瑅
+資
+縤
+陣
+蕟
+栢
+韻
+祼
+恁
+伢
+謝
+劃
+涑
+總
+衖
+踺
+砋
+凉
+籃
+駿
+苼
+瘋
+昽
+紡
+驊
+腎
+﹗
+響
+杋
+剛
+嚴
+禪
+歓
+槍
+傘
+檸
+檫
+炣
+勢
+鏜
+鎢
+銑
+尐
+減
+奪
+惡
+θ
+僮
+婭
+臘
+ū
+ì
+殻
+鉄
+∑
+蛲
+焼
+緖
+續
+紹
+懮

deepdoc/models/rec.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c7cf60de2afd728d512f4190cf37455092b45f06175365c6fc58d8cd7e2a68b
+size 10826336

deepdoc/models/tsr.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04c14c3e41802450a1f437a3865ce1a3186046262ea4d75c8975289687a43223
+size 12243020

deepdoc/vision/__init__.py

@@ -0,0 +1,3 @@
+from .ragFlow import RagFlow 
+
+__all__ = ['ragFlow']

deepdoc/vision/ocr.res

@@ -0,0 +1,6623 @@
+'
+疗
+绚
+诚
+娇
+溜
+题
+贿
+者
+廖
+更
+纳
+加
+奉
+公
+一
+就
+汴
+计
+与
+路
+房
+原
+妇
+2
+0
+8
+-
+7
+其
+>
+:
+]
+,
+，
+骑
+刈
+全
+消
+昏
+傈
+安
+久
+钟
+嗅
+不
+影
+处
+驽
+蜿
+资
+关
+椤
+地
+瘸
+专
+问
+忖
+票
+嫉
+炎
+韵
+要
+月
+田
+节
+陂
+鄙
+捌
+备
+拳
+伺
+眼
+网
+盎
+大
+傍
+心
+东
+愉
+汇
+蹿
+科
+每
+业
+里
+航
+晏
+字
+平
+录
+先
+1
+3
+彤
+鲶
+产
+稍
+督
+腴
+有
+象
+岳
+注
+绍
+在
+泺
+文
+定
+核
+名
+水
+过
+理
+让
+偷
+率
+等
+这
+发
+”
+为
+含
+肥
+酉
+相
+鄱
+七
+编
+猥
+锛
+日
+镀
+蒂
+掰
+倒
+辆
+栾
+栗
+综
+涩
+州
+雌
+滑
+馀
+了
+机
+块
+司
+宰
+甙
+兴
+矽
+抚
+保
+用
+沧
+秩
+如
+收
+息
+滥
+页
+疑
+埠
+!
+！
+姥
+异
+橹
+钇
+向
+下
+跄
+的
+椴
+沫
+国
+绥
+獠
+报
+开
+民
+蜇
+何
+分
+凇
+长
+讥
+藏
+掏
+施
+羽
+中
+讲
+派
+嘟
+人
+提
+浼
+间
+世
+而
+古
+多
+倪
+唇
+饯
+控
+庚
+首
+赛
+蜓
+味
+断
+制
+觉
+技
+替
+艰
+溢
+潮
+夕
+钺
+外
+摘
+枋
+动
+双
+单
+啮
+户
+枇
+确
+锦
+曜
+杜
+或
+能
+效
+霜
+盒
+然
+侗
+电
+晁
+放
+步
+鹃
+新
+杖
+蜂
+吒
+濂
+瞬
+评
+总
+隍
+对
+独
+合
+也
+是
+府
+青
+天
+诲
+墙
+组
+滴
+级
+邀
+帘
+示
+已
+时
+骸
+仄
+泅
+和
+遨
+店
+雇
+疫
+持
+巍
+踮
+境
+只
+亨
+目
+鉴
+崤
+闲
+体
+泄
+杂
+作
+般
+轰
+化
+解
+迂
+诿
+蛭
+璀
+腾
+告
+版
+服
+省
+师
+小
+规
+程
+线
+海
+办
+引
+二
+桧
+牌
+砺
+洄
+裴
+修
+图
+痫
+胡
+许
+犊
+事
+郛
+基
+柴
+呼
+食
+研
+奶
+律
+蛋
+因
+葆
+察
+戏
+褒
+戒
+再
+李
+骁
+工
+貂
+油
+鹅
+章
+啄
+休
+场
+给
+睡
+纷
+豆
+器
+捎
+说
+敏
+学
+会
+浒
+设
+诊
+格
+廓
+查
+来
+霓
+室
+溆
+￠
+诡
+寥
+焕
+舜
+柒
+狐
+回
+戟
+砾
+厄
+实
+翩
+尿
+五
+入
+径
+惭
+喹
+股
+宇
+篝
+|
+;
+美
+期
+云
+九
+祺
+扮
+靠
+锝
+槌
+系
+企
+酰
+阊
+暂
+蚕
+忻
+豁
+本
+羹
+执
+条
+钦
+H
+獒
+限
+进
+季
+楦
+于
+芘
+玖
+铋
+茯
+未
+答
+粘
+括
+样
+精
+欠
+矢
+甥
+帷
+嵩
+扣
+令
+仔
+风
+皈
+行
+支
+部
+蓉
+刮
+站
+蜡
+救
+钊
+汗
+松
+嫌
+成
+可
+.
+鹤
+院
+从
+交
+政
+怕
+活
+调
+球
+局
+验
+髌
+第
+韫
+谗
+串
+到
+圆
+年
+米
+/
+*
+友
+忿
+检
+区
+看
+自
+敢
+刃
+个
+兹
+弄
+流
+留
+同
+没
+齿
+星
+聆
+轼
+湖
+什
+三
+建
+蛔
+儿
+椋
+汕
+震
+颧
+鲤
+跟
+力
+情
+璺
+铨
+陪
+务
+指
+族
+训
+滦
+鄣
+濮
+扒
+商
+箱
+十
+召
+慷
+辗
+所
+莞
+管
+护
+臭
+横
+硒
+嗓
+接
+侦
+六
+露
+党
+馋
+驾
+剖
+高
+侬
+妪
+幂
+猗
+绺
+骐
+央
+酐
+孝
+筝
+课
+徇
+缰
+门
+男
+西
+项
+句
+谙
+瞒
+秃
+篇
+教
+碲
+罚
+声
+呐
+景
+前
+富
+嘴
+鳌
+稀
+免
+朋
+啬
+睐
+去
+赈
+鱼
+住
+肩
+愕
+速
+旁
+波
+厅
+健
+茼
+厥
+鲟
+谅
+投
+攸
+炔
+数
+方
+击
+呋
+谈
+绩
+别
+愫
+僚
+躬
+鹧
+胪
+炳
+招
+喇
+膨
+泵
+蹦
+毛
+结
+5
+4
+谱
+识
+陕
+粽
+婚
+拟
+构
+且
+搜
+任
+潘
+比
+郢
+妨
+醪
+陀
+桔
+碘
+扎
+选
+哈
+骷
+楷
+亿
+明
+缆
+脯
+监
+睫
+逻
+婵
+共
+赴
+淝
+凡
+惦
+及
+达
+揖
+谩
+澹
+减
+焰
+蛹
+番
+祁
+柏
+员
+禄
+怡
+峤
+龙
+白
+叽
+生
+闯
+起
+细
+装
+谕
+竟
+聚
+钙
+上
+导
+渊
+按
+艾
+辘
+挡
+耒
+盹
+饪
+臀
+记
+邮
+蕙
+受
+各
+医
+搂
+普
+滇
+朗
+茸
+带
+翻
+酚
+(
+光
+堤
+墟
+蔷
+万
+幻
+〓
+瑙
+辈
+昧
+盏
+亘
+蛀
+吉
+铰
+请
+子
+假
+闻
+税
+井
+诩
+哨
+嫂
+好
+面
+琐
+校
+馊
+鬣
+缂
+营
+访
+炖
+占
+农
+缀
+否
+经
+钚
+棵
+趟
+张
+亟
+吏
+茶
+谨
+捻
+论
+迸
+堂
+玉
+信
+吧
+瞠
+乡
+姬
+寺
+咬
+溏
+苄
+皿
+意
+赉
+宝
+尔
+钰
+艺
+特
+唳
+踉
+都
+荣
+倚
+登
+荐
+丧
+奇
+涵
+批
+炭
+近
+符
+傩
+感
+道
+着
+菊
+虹
+仲
+众
+懈
+濯
+颞
+眺
+南
+释
+北
+缝
+标
+既
+茗
+整
+撼
+迤
+贲
+挎
+耱
+拒
+某
+妍
+卫
+哇
+英
+矶
+藩
+治
+他
+元
+领
+膜
+遮
+穗
+蛾
+飞
+荒
+棺
+劫
+么
+市
+火
+温
+拈
+棚
+洼
+转
+��
+奕
+卸
+迪
+伸
+泳
+斗
+邡
+侄
+涨
+屯
+萋
+胭
+氡
+崮
+枞
+惧
+冒
+彩
+斜
+手
+豚
+随
+旭
+淑
+妞
+形
+菌
+吲
+沱
+争
+驯
+歹
+挟
+兆
+柱
+传
+至
+包
+内
+响
+临
+红
+功
+弩
+衡
+寂
+禁
+老
+棍
+耆
+渍
+织
+害
+氵
+渑
+布
+载
+靥
+嗬
+虽
+苹
+咨
+娄
+库
+雉
+榜
+帜
+嘲
+套
+瑚
+亲
+簸
+欧
+边
+6
+腿
+旮
+抛
+吹
+瞳
+得
+镓
+梗
+厨
+继
+漾
+愣
+憨
+士
+策
+窑
+抑
+躯
+襟
+脏
+参
+贸
+言
+干
+绸
+鳄
+穷
+藜
+音
+折
+详
+)
+举
+悍
+甸
+癌
+黎
+谴
+死
+罩
+迁
+寒
+驷
+袖
+媒
+蒋
+掘
+模
+纠
+恣
+观
+祖
+蛆
+碍
+位
+稿
+主
+澧
+跌
+筏
+京
+锏
+帝
+贴
+证
+糠
+才
+黄
+鲸
+略
+炯
+饱
+四
+出
+园
+犀
+牧
+容
+汉
+杆
+浈
+汰
+瑷
+造
+虫
+瘩
+怪
+驴
+济
+应
+花
+沣
+谔
+夙
+旅
+价
+矿
+以
+考
+s
+u
+呦
+晒
+巡
+茅
+准
+肟
+瓴
+詹
+仟
+褂
+译
+桌
+混
+宁
+怦
+郑
+抿
+些
+余
+鄂
+饴
+攒
+珑
+群
+阖
+岔
+琨
+藓
+预
+环
+洮
+岌
+宀
+杲
+瀵
+最
+常
+囡
+周
+踊
+女
+鼓
+袭
+喉
+简
+范
+薯
+遐
+疏
+粱
+黜
+禧
+法
+箔
+斤
+遥
+汝
+奥
+直
+贞
+撑
+置
+绱
+集
+她
+馅
+逗
+钧
+橱
+魉
+[
+恙
+躁
+唤
+9
+旺
+膘
+待
+脾
+惫
+购
+吗
+依
+盲
+度
+瘿
+蠖
+俾
+之
+镗
+拇
+鲵
+厝
+簧
+续
+款
+展
+啃
+表
+剔
+品
+钻
+腭
+损
+清
+锶
+统
+涌
+寸
+滨
+贪
+链
+吠
+冈
+伎
+迥
+咏
+吁
+览
+防
+迅
+失
+汾
+阔
+逵
+绀
+蔑
+列
+川
+凭
+努
+熨
+揪
+利
+俱
+绉
+抢
+鸨
+我
+即
+责
+膦
+易
+毓
+鹊
+刹
+玷
+岿
+空
+嘞
+绊
+排
+术
+估
+锷
+违
+们
+苟
+铜
+播
+肘
+件
+烫
+审
+鲂
+广
+像
+铌
+惰
+铟
+巳
+胍
+鲍
+康
+憧
+色
+恢
+想
+拷
+尤
+疳
+知
+S
+Y
+F
+D
+A
+峄
+裕
+帮
+握
+搔
+氐
+氘
+难
+墒
+沮
+雨
+叁
+缥
+悴
+藐
+湫
+娟
+苑
+稠
+颛
+簇
+后
+阕
+闭
+蕤
+缚
+怎
+佞
+码
+嘤
+蔡
+痊
+舱
+螯
+帕
+赫
+昵
+升
+烬
+岫
+、
+疵
+蜻
+髁
+蕨
+隶
+烛
+械
+丑
+盂
+梁
+强
+鲛
+由
+拘
+揉
+劭
+龟
+撤
+钩
+呕
+孛
+费
+妻
+漂
+求
+阑
+崖
+秤
+甘
+通
+深
+补
+赃
+坎
+床
+啪
+承
+吼
+量
+暇
+钼
+烨
+阂
+擎
+脱
+逮
+称
+P
+神
+属
+矗
+华
+届
+狍
+葑
+汹
+育
+患
+窒
+蛰
+佼
+静
+槎
+运
+鳗
+庆
+逝
+曼
+疱
+克
+代
+官
+此
+麸
+耧
+蚌
+晟
+例
+础
+榛
+副
+测
+唰
+缢
+迹
+灬
+霁
+身
+岁
+赭
+扛
+又
+菡
+乜
+雾
+板
+读
+陷
+徉
+贯
+郁
+虑
+变
+钓
+菜
+圾
+现
+琢
+式
+乐
+维
+渔
+浜
+左
+吾
+脑
+钡
+警
+T
+啵
+拴
+偌
+漱
+湿
+硕
+止
+骼
+魄
+积
+燥
+联
+踢
+玛
+则
+窿
+见
+振
+畿
+送
+班
+钽
+您
+赵
+刨
+印
+讨
+踝
+籍
+谡
+舌
+崧
+汽
+蔽
+沪
+酥
+绒
+怖
+财
+帖
+肱
+私
+莎
+勋
+羔
+霸
+励
+哼
+帐
+将
+帅
+渠
+纪
+婴
+娩
+岭
+厘
+滕
+吻
+伤
+坝
+冠
+戊
+隆
+瘁
+介
+涧
+物
+黍
+并
+姗
+奢
+蹑
+掣
+垸
+锴
+命
+箍
+捉
+病
+辖
+琰
+眭
+迩
+艘
+绌
+繁
+寅
+若
+毋
+思
+诉
+类
+诈
+燮
+轲
+酮
+狂
+重
+反
+职
+筱
+县
+委
+磕
+绣
+奖
+晋
+濉
+志
+徽
+肠
+呈
+獐
+坻
+口
+片
+碰
+几
+村
+柿
+劳
+料
+获
+亩
+惕
+晕
+厌
+号
+罢
+池
+正
+鏖
+煨
+家
+棕
+复
+尝
+懋
+蜥
+锅
+岛
+扰
+队
+坠
+瘾
+钬
+@
+卧
+疣
+镇
+譬
+冰
+彷
+频
+黯
+据
+垄
+采
+八
+缪
+瘫
+型
+熹
+砰
+楠
+襁
+箐
+但
+嘶
+绳
+啤
+拍
+盥
+穆
+傲
+洗
+盯
+塘
+怔
+筛
+丿
+台
+恒
+喂
+葛
+永
+￥
+烟
+酒
+桦
+书
+砂
+蚝
+缉
+态
+瀚
+袄
+圳
+轻
+蛛
+超
+榧
+遛
+姒
+奘
+铮
+右
+荽
+望
+偻
+卡
+丶
+氰
+附
+做
+革
+索
+戚
+坨
+桷
+唁
+垅
+榻
+岐
+偎
+坛
+莨
+山
+殊
+微
+骇
+陈
+爨
+推
+嗝
+驹
+澡
+藁
+呤
+卤
+嘻
+糅
+逛
+侵
+郓
+酌
+德
+摇
+※
+鬃
+被
+慨
+殡
+羸
+昌
+泡
+戛
+鞋
+河
+宪
+沿
+玲
+鲨
+翅
+哽
+源
+铅
+语
+照
+邯
+址
+荃
+佬
+顺
+鸳
+町
+霭
+睾
+瓢
+夸
+椁
+晓
+酿
+痈
+咔
+侏
+券
+噎
+湍
+签
+嚷
+离
+午
+尚
+社
+锤
+背
+孟
+使
+浪
+缦
+潍
+鞅
+军
+姹
+驶
+笑
+鳟
+鲁
+》
+孽
+钜
+绿
+洱
+礴
+焯
+椰
+颖
+囔
+乌
+孔
+巴
+互
+性
+椽
+哞
+聘
+昨
+早
+暮
+胶
+炀
+隧
+��
+彗
+昝
+铁
+呓
+氽
+藉
+喔
+癖
+瑗
+姨
+权
+胱
+韦
+堑
+蜜
+酋
+楝
+砝
+毁
+靓
+歙
+锲
+究
+屋
+喳
+骨
+辨
+碑
+武
+鸠
+宫
+辜
+烊
+适
+坡
+殃
+培
+佩
+供
+走
+蜈
+迟
+翼
+况
+姣
+凛
+浔
+吃
+飘
+债
+犟
+金
+促
+苛
+崇
+坂
+莳
+畔
+绂
+兵
+蠕
+斋
+根
+砍
+亢
+欢
+恬
+崔
+剁
+餐
+榫
+快
+扶
+‖
+濒
+缠
+鳜
+当
+彭
+驭
+浦
+篮
+昀
+锆
+秸
+钳
+弋
+娣
+瞑
+夷
+龛
+苫
+拱
+致
+%
+嵊
+障
+隐
+弑
+初
+娓
+抉
+汩
+累
+蓖
+"
+唬
+助
+苓
+昙
+押
+毙
+破
+城
+郧
+逢
+嚏
+獭
+瞻
+溱
+婿
+赊
+跨
+恼
+璧
+萃
+姻
+貉
+灵
+炉
+密
+氛
+陶
+砸
+谬
+衔
+点
+琛
+沛
+枳
+层
+岱
+诺
+脍
+榈
+埂
+征
+冷
+裁
+打
+蹴
+素
+瘘
+逞
+蛐
+聊
+激
+腱
+萘
+踵
+飒
+蓟
+吆
+取
+咙
+簋
+涓
+矩
+曝
+挺
+揣
+座
+你
+史
+舵
+焱
+尘
+苏
+笈
+脚
+溉
+榨
+诵
+樊
+邓
+焊
+义
+庶
+儋
+蟋
+蒲
+赦
+呷
+杞
+诠
+豪
+还
+试
+颓
+茉
+太
+除
+紫
+逃
+痴
+草
+充
+鳕
+珉
+祗
+墨
+渭
+烩
+蘸
+慕
+璇
+镶
+穴
+嵘
+恶
+骂
+险
+绋
+幕
+碉
+肺
+戳
+刘
+潞
+秣
+纾
+潜
+銮
+洛
+须
+罘
+销
+瘪
+汞
+兮
+屉
+r
+林
+厕
+质
+探
+划
+狸
+殚
+善
+煊
+烹
+〒
+锈
+逯
+宸
+辍
+泱
+柚
+袍
+远
+蹋
+嶙
+绝
+峥
+娥
+缍
+雀
+徵
+认
+镱
+谷
+=
+贩
+勉
+撩
+鄯
+斐
+洋
+非
+祚
+泾
+诒
+饿
+撬
+威
+晷
+搭
+芍
+锥
+笺
+蓦
+候
+琊
+档
+礁
+沼
+卵
+荠
+忑
+朝
+凹
+瑞
+头
+仪
+弧
+孵
+畏
+铆
+突
+衲
+车
+浩
+气
+茂
+悖
+厢
+枕
+酝
+戴
+湾
+邹
+飚
+攘
+锂
+写
+宵
+翁
+岷
+无
+喜
+丈
+挑
+嗟
+绛
+殉
+议
+槽
+具
+醇
+淞
+笃
+郴
+阅
+饼
+底
+壕
+砚
+弈
+询
+缕
+庹
+翟
+零
+筷
+暨
+舟
+闺
+甯
+撞
+麂
+茌
+蔼
+很
+珲
+捕
+棠
+角
+阉
+媛
+娲
+诽
+剿
+尉
+爵
+睬
+韩
+诰
+匣
+危
+糍
+镯
+立
+浏
+阳
+少
+盆
+舔
+擘
+匪
+申
+尬
+铣
+旯
+抖
+赘
+瓯
+居
+ˇ
+哮
+游
+锭
+茏
+歌
+坏
+甚
+秒
+舞
+沙
+仗
+劲
+潺
+阿
+燧
+郭
+嗖
+霏
+忠
+材
+奂
+耐
+跺
+砀
+输
+岖
+媳
+氟
+极
+摆
+灿
+今
+扔
+腻
+枝
+奎
+药
+熄
+吨
+话
+q
+额
+慑
+嘌
+协
+喀
+壳
+埭
+视
+著
+於
+愧
+陲
+翌
+峁
+颅
+佛
+腹
+聋
+侯
+咎
+叟
+秀
+颇
+存
+较
+罪
+哄
+岗
+扫
+栏
+钾
+羌
+己
+璨
+枭
+霉
+煌
+涸
+衿
+键
+镝
+益
+岢
+奏
+连
+夯
+睿
+冥
+均
+糖
+狞
+蹊
+稻
+爸
+刿
+胥
+煜
+丽
+肿
+璃
+掸
+跚
+灾
+垂
+樾
+濑
+乎
+莲
+窄
+犹
+撮
+战
+馄
+软
+络
+显
+鸢
+胸
+宾
+妲
+恕
+埔
+蝌
+份
+遇
+巧
+瞟
+粒
+恰
+剥
+桡
+博
+讯
+凯
+堇
+阶
+滤
+卖
+斌
+骚
+彬
+兑
+磺
+樱
+舷
+两
+娱
+福
+仃
+差
+找
+桁
+÷
+净
+把
+阴
+污
+戬
+雷
+碓
+蕲
+楚
+罡
+焖
+抽
+妫
+咒
+仑
+闱
+尽
+邑
+菁
+爱
+贷
+沥
+鞑
+牡
+嗉
+崴
+骤
+塌
+嗦
+订
+拮
+滓
+捡
+锻
+次
+坪
+杩
+臃
+箬
+融
+珂
+鹗
+宗
+枚
+降
+鸬
+妯
+阄
+堰
+盐
+毅
+必
+杨
+崃
+俺
+甬
+状
+莘
+货
+耸
+菱
+腼
+铸
+唏
+痤
+孚
+澳
+懒
+溅
+翘
+疙
+杷
+淼
+缙
+骰
+喊
+悉
+砻
+坷
+艇
+赁
+界
+谤
+纣
+宴
+晃
+茹
+归
+饭
+梢
+铡
+街
+抄
+肼
+鬟
+苯
+颂
+撷
+戈
+炒
+咆
+茭
+瘙
+负
+仰
+客
+琉
+铢
+封
+卑
+珥
+椿
+镧
+窨
+鬲
+寿
+御
+袤
+铃
+萎
+砖
+餮
+脒
+裳
+肪
+孕
+嫣
+馗
+嵇
+恳
+氯
+江
+石
+褶
+冢
+祸
+阻
+狈
+羞
+银
+靳
+透
+咳
+叼
+敷
+芷
+啥
+它
+瓤
+兰
+痘
+懊
+逑
+肌
+往
+捺
+坊
+甩
+呻
+〃
+沦
+忘
+膻
+祟
+菅
+剧
+崆
+智
+坯
+臧
+霍
+墅
+攻
+眯
+倘
+拢
+骠
+铐
+庭
+岙
+瓠
+′
+缺
+泥
+迢
+捶
+?
+？
+郏
+喙
+掷
+沌
+纯
+秘
+种
+听
+绘
+固
+螨
+团
+香
+盗
+妒
+埚
+蓝
+拖
+旱
+荞
+铀
+血
+遏
+汲
+辰
+叩
+拽
+幅
+硬
+惶
+桀
+漠
+措
+泼
+唑
+齐
+肾
+念
+酱
+虚
+屁
+耶
+旗
+砦
+闵
+婉
+馆
+拭
+绅
+韧
+忏
+窝
+醋
+葺
+顾
+辞
+倜
+堆
+辋
+逆
+玟
+贱
+疾
+董
+惘
+倌
+锕
+淘
+嘀
+莽
+俭
+笏
+绑
+鲷
+杈
+择
+蟀
+粥
+嗯
+驰
+逾
+案
+谪
+褓
+胫
+哩
+昕
+颚
+鲢
+绠
+躺
+鹄
+崂
+儒
+俨
+丝
+尕
+泌
+啊
+萸
+彰
+幺
+吟
+骄
+苣
+弦
+脊
+瑰
+〈
+诛
+镁
+析
+闪
+剪
+侧
+哟
+框
+螃
+守
+嬗
+燕
+狭
+铈
+缮
+概
+迳
+痧
+鲲
+俯
+售
+笼
+痣
+扉
+挖
+满
+咋
+援
+邱
+扇
+歪
+便
+玑
+绦
+峡
+蛇
+叨
+〖
+泽
+胃
+斓
+喋
+怂
+坟
+猪
+该
+蚬
+炕
+弥
+赞
+棣
+晔
+娠
+挲
+狡
+创
+疖
+铕
+镭
+稷
+挫
+弭
+啾
+翔
+粉
+履
+苘
+哦
+楼
+秕
+铂
+土
+锣
+瘟
+挣
+栉
+习
+享
+桢
+袅
+磨
+桂
+谦
+延
+坚
+蔚
+噗
+署
+谟
+猬
+钎
+恐
+嬉
+雒
+倦
+衅
+亏
+璩
+睹
+刻
+殿
+王
+算
+雕
+麻
+丘
+柯
+骆
+丸
+塍
+谚
+添
+鲈
+垓
+桎
+蚯
+芥
+予
+飕
+镦
+谌
+窗
+醚
+菀
+亮
+搪
+莺
+蒿
+羁
+足
+J
+真
+轶
+悬
+衷
+靛
+翊
+掩
+哒
+炅
+掐
+冼
+妮
+l
+谐
+稚
+荆
+擒
+犯
+陵
+虏
+浓
+崽
+刍
+陌
+傻
+孜
+千
+靖
+演
+矜
+钕
+煽
+杰
+酗
+渗
+伞
+栋
+俗
+泫
+戍
+罕
+沾
+疽
+灏
+煦
+芬
+磴
+叱
+阱
+榉
+湃
+蜀
+叉
+醒
+彪
+租
+郡
+篷
+屎
+良
+垢
+隗
+弱
+陨
+峪
+砷
+掴
+颁
+胎
+雯
+绵
+贬
+沐
+撵
+隘
+篙
+暖
+曹
+陡
+栓
+填
+臼
+彦
+瓶
+琪
+潼
+哪
+鸡
+摩
+啦
+俟
+锋
+域
+耻
+蔫
+疯
+纹
+撇
+毒
+绶
+痛
+酯
+忍
+爪
+赳
+歆
+嘹
+辕
+烈
+册
+朴
+钱
+吮
+毯
+癜
+娃
+谀
+邵
+厮
+炽
+璞
+邃
+丐
+追
+词
+瓒
+忆
+轧
+芫
+谯
+喷
+弟
+半
+冕
+裙
+掖
+墉
+绮
+寝
+苔
+势
+顷
+褥
+切
+衮
+君
+佳
+嫒
+蚩
+霞
+佚
+洙
+逊
+镖
+暹
+唛
+&
+殒
+顶
+碗
+獗
+轭
+铺
+蛊
+废
+恹
+汨
+崩
+珍
+那
+杵
+曲
+纺
+夏
+薰
+傀
+闳
+淬
+姘
+舀
+拧
+卷
+楂
+恍
+讪
+厩
+寮
+篪
+赓
+乘
+灭
+盅
+鞣
+沟
+慎
+挂
+饺
+鼾
+杳
+树
+缨
+丛
+絮
+娌
+臻
+嗳
+篡
+侩
+述
+衰
+矛
+圈
+蚜
+匕
+筹
+匿
+濞
+晨
+叶
+骋
+郝
+挚
+蚴
+滞
+增
+侍
+描
+瓣
+吖
+嫦
+蟒
+匾
+圣
+赌
+毡
+癞
+恺
+百
+曳
+需
+篓
+肮
+庖
+帏
+卿
+驿
+遗
+蹬
+鬓
+骡
+歉
+芎
+胳
+屐
+禽
+烦
+晌
+寄
+媾
+狄
+翡
+苒
+船
+廉
+终
+痞
+殇
+々
+畦
+饶
+改
+拆
+悻
+萄
+￡
+瓿
+乃
+訾
+桅
+匮
+溧
+拥
+纱
+铍
+骗
+蕃
+龋
+缬
+父
+佐
+疚
+栎
+醍
+掳
+蓄
+x
+惆
+颜
+鲆
+榆
+〔
+猎
+敌
+暴
+谥
+鲫
+贾
+罗
+玻
+缄
+扦
+芪
+癣
+落
+徒
+臾
+恿
+猩
+托
+邴
+肄
+牵
+春
+陛
+耀
+刊
+拓
+蓓
+邳
+堕
+寇
+枉
+淌
+啡
+湄
+兽
+酷
+萼
+碚
+濠
+萤
+夹
+旬
+戮
+梭
+琥
+椭
+昔
+勺
+蜊
+绐
+晚
+孺
+僵
+宣
+摄
+冽
+旨
+萌
+忙
+蚤
+眉
+噼
+蟑
+付
+契
+瓜
+悼
+颡
+壁
+曾
+窕
+颢
+澎
+仿
+俑
+浑
+嵌
+浣
+乍
+碌
+褪
+乱
+蔟
+隙
+玩
+剐
+葫
+箫
+纲
+围
+伐
+决
+伙
+漩
+瑟
+刑
+肓
+镳
+缓
+蹭
+氨
+皓
+典
+畲
+坍
+铑
+檐
+塑
+洞
+倬
+储
+胴
+淳
+戾
+吐
+灼
+惺
+妙
+毕
+珐
+缈
+虱
+盖
+羰
+鸿
+磅
+谓
+髅
+娴
+苴
+唷
+蚣
+霹
+抨
+贤
+唠
+犬
+誓
+逍
+庠
+逼
+麓
+籼
+釉
+呜
+碧
+秧
+氩
+摔
+霄
+穸
+纨
+辟
+妈
+映
+完
+牛
+缴
+嗷
+炊
+恩
+荔
+茆
+掉
+紊
+慌
+莓
+羟
+阙
+萁
+磐
+另
+蕹
+辱
+鳐
+湮
+吡
+吩
+唐
+睦
+垠
+舒
+圜
+冗
+瞿
+溺
+芾
+囱
+匠
+僳
+汐
+菩
+饬
+漓
+黑
+霰
+浸
+濡
+窥
+毂
+蒡
+兢
+驻
+鹉
+芮
+诙
+迫
+雳
+厂
+忐
+臆
+猴
+鸣
+蚪
+栈
+箕
+羡
+渐
+莆
+捍
+眈
+哓
+趴
+蹼
+埕
+嚣
+骛
+宏
+淄
+斑
+噜
+严
+瑛
+垃
+椎
+诱
+压
+庾
+绞
+焘
+廿
+抡
+迄
+棘
+夫
+纬
+锹
+眨
+瞌
+侠
+脐
+竞
+瀑
+孳
+骧
+遁
+姜
+颦
+荪
+滚
+萦
+伪
+逸
+粳
+爬
+锁
+矣
+役
+趣
+洒
+颔
+诏
+逐
+奸
+甭
+惠
+攀
+蹄
+泛
+尼
+拼
+阮
+鹰
+亚
+颈
+惑
+勒
+〉
+际
+肛
+爷
+刚
+钨
+丰
+养
+冶
+鲽
+辉
+蔻
+画
+覆
+皴
+妊
+麦
+返
+醉
+皂
+擀
+〗
+酶
+凑
+粹
+悟
+诀
+硖
+港
+卜
+z
+杀
+涕
+±
+舍
+铠
+抵
+弛
+段
+敝
+镐
+奠
+拂
+轴
+跛
+袱
+e
+t
+沉
+菇
+俎
+薪
+峦
+秭
+蟹
+历
+盟
+菠
+寡
+液
+肢
+喻
+染
+裱
+悱
+抱
+氙
+赤
+捅
+猛
+跑
+氮
+谣
+仁
+尺
+辊
+窍
+烙
+衍
+架
+擦
+倏
+璐
+瑁
+币
+楞
+胖
+夔
+趸
+邛
+惴
+饕
+虔
+蝎
+§
+哉
+贝
+宽
+辫
+炮
+扩
+饲
+籽
+魏
+菟
+锰
+伍
+猝
+末
+琳
+哚
+蛎
+邂
+呀
+姿
+鄞
+却
+歧
+仙
+恸
+椐
+森
+牒
+寤
+袒
+婆
+虢
+雅
+钉
+朵
+贼
+欲
+苞
+寰
+故
+龚
+坭
+嘘
+咫
+礼
+硷
+兀
+睢
+汶
+’
+铲
+烧
+绕
+诃
+浃
+钿
+哺
+柜
+讼
+颊
+璁
+腔
+洽
+咐
+脲
+簌
+筠
+镣
+玮
+鞠
+谁
+兼
+姆
+挥
+梯
+蝴
+谘
+漕
+刷
+躏
+宦
+弼
+b
+垌
+劈
+麟
+莉
+揭
+笙
+渎
+仕
+嗤
+仓
+配
+怏
+抬
+错
+泯
+镊
+孰
+猿
+邪
+仍
+秋
+鼬
+壹
+歇
+吵
+炼
+<
+尧
+射
+柬
+廷
+胧
+霾
+凳
+隋
+肚
+浮
+梦
+祥
+株
+堵
+退
+L
+鹫
+跎
+凶
+毽
+荟
+炫
+栩
+玳
+甜
+沂
+鹿
+顽
+伯
+爹
+赔
+蛴
+徐
+匡
+欣
+狰
+缸
+雹
+蟆
+疤
+默
+沤
+啜
+痂
+衣
+禅
+w
+i
+h
+辽
+葳
+黝
+钗
+停
+沽
+棒
+馨
+颌
+肉
+吴
+硫
+悯
+劾
+娈
+马
+啧
+吊
+悌
+镑
+峭
+帆
+瀣
+涉
+咸
+疸
+滋
+泣
+翦
+拙
+癸
+钥
+蜒
++
+尾
+庄
+凝
+泉
+婢
+渴
+谊
+乞
+陆
+锉
+糊
+鸦
+淮
+I
+B
+N
+晦
+弗
+乔
+庥
+葡
+尻
+席
+橡
+傣
+渣
+拿
+惩
+麋
+斛
+缃
+矮
+蛏
+岘
+鸽
+姐
+膏
+催
+奔
+镒
+喱
+蠡
+摧
+钯
+胤
+柠
+拐
+璋
+鸥
+卢
+荡
+倾
+^
+_
+珀
+逄
+萧
+塾
+掇
+贮
+笆
+聂
+圃
+冲
+嵬
+M
+滔
+笕
+值
+炙
+偶
+蜱
+搐
+梆
+汪
+蔬
+腑
+鸯
+蹇
+敞
+绯
+仨
+祯
+谆
+梧
+糗
+鑫
+啸
+豺
+囹
+猾
+巢
+柄
+瀛
+筑
+踌
+沭
+暗
+苁
+鱿
+蹉
+脂
+蘖
+牢
+热
+木
+吸
+溃
+宠
+序
+泞
+偿
+拜
+檩
+厚
+朐
+毗
+螳
+吞
+媚
+朽
+担
+蝗
+橘
+畴
+祈
+糟
+盱
+隼
+郜
+惜
+珠
+裨
+铵
+焙
+琚
+唯
+咚
+噪
+骊
+丫
+滢
+勤
+棉
+呸
+咣
+淀
+隔
+蕾
+窈
+饨
+挨
+煅
+短
+匙
+粕
+镜
+赣
+撕
+墩
+酬
+馁
+豌
+颐
+抗
+酣
+氓
+佑
+搁
+哭
+递
+耷
+涡
+桃
+贻
+碣
+截
+瘦
+昭
+镌
+蔓
+氚
+甲
+猕
+蕴
+蓬
+散
+拾
+纛
+狼
+猷
+铎
+埋
+旖
+矾
+讳
+囊
+糜
+迈
+粟
+蚂
+紧
+鲳
+瘢
+栽
+稼
+羊
+锄
+斟
+睁
+桥
+瓮
+蹙
+祉
+醺
+鼻
+昱
+剃
+跳
+篱
+跷
+蒜
+翎
+宅
+晖
+嗑
+壑
+峻
+癫
+屏
+狠
+陋
+袜
+途
+憎
+祀
+莹
+滟
+佶
+溥
+臣
+约
+盛
+峰
+磁
+慵
+婪
+拦
+莅
+朕
+鹦
+粲
+裤
+哎
+疡
+嫖
+琵
+窟
+堪
+谛
+嘉
+儡
+鳝
+斩
+郾
+驸
+酊
+妄
+胜
+贺
+徙
+傅
+噌
+钢
+栅
+庇
+恋
+匝
+巯
+邈
+尸
+锚
+粗
+佟
+蛟
+薹
+纵
+蚊
+郅
+绢
+锐
+苗
+俞
+篆
+淆
+膀
+鲜
+煎
+诶
+秽
+寻
+涮
+刺
+怀
+噶
+巨
+褰
+魅
+灶
+灌
+桉
+藕
+谜
+舸
+薄
+搀
+恽
+借
+牯
+痉
+渥
+愿
+亓
+耘
+杠
+柩
+锔
+蚶
+钣
+珈
+喘
+蹒
+幽
+赐
+稗
+晤
+莱
+泔
+扯
+肯
+菪
+裆
+腩
+豉
+疆
+骜
+腐
+倭
+珏
+唔
+粮
+亡
+润
+慰
+伽
+橄
+玄
+誉
+醐
+胆
+龊
+粼
+塬
+陇
+彼
+削
+嗣
+绾
+芽
+妗
+垭
+瘴
+爽
+薏
+寨
+龈
+泠
+弹
+赢
+漪
+猫
+嘧
+涂
+恤
+圭
+茧
+烽
+屑
+痕
+巾
+赖
+荸
+凰
+腮
+畈
+亵
+蹲
+偃
+苇
+澜
+艮
+换
+骺
+烘
+苕
+梓
+颉
+肇
+哗
+悄
+氤
+涠
+葬
+屠
+鹭
+植
+竺
+佯
+诣
+鲇
+瘀
+鲅
+邦
+移
+滁
+冯
+耕
+癔
+戌
+茬
+沁
+巩
+悠
+湘
+洪
+痹
+锟
+循
+谋
+腕
+鳃
+钠
+捞
+焉
+迎
+碱
+伫
+急
+榷
+奈
+邝
+卯
+辄
+皲
+卟
+醛
+畹
+忧
+稳
+雄
+昼
+缩
+阈
+睑
+扌
+耗
+曦
+涅
+捏
+瞧
+邕
+淖
+漉
+铝
+耦
+禹
+湛
+喽
+莼
+琅
+诸
+苎
+纂
+硅
+始
+嗨
+傥
+燃
+臂
+赅
+嘈
+呆
+贵
+屹
+壮
+肋
+亍
+蚀
+卅
+豹
+腆
+邬
+迭
+浊
+}
+童
+螂
+捐
+圩
+勐
+触
+寞
+汊
+壤
+荫
+膺
+渌
+芳
+懿
+遴
+螈
+泰
+蓼
+蛤
+茜
+舅
+枫
+朔
+膝
+眙
+避
+梅
+判
+鹜
+璜
+牍
+缅
+垫
+藻
+黔
+侥
+惚
+懂
+踩
+腰
+腈
+札
+丞
+唾
+慈
+顿
+摹
+荻
+琬
+~
+斧
+沈
+滂
+胁
+胀
+幄
+莜
+Z
+匀
+鄄
+掌
+绰
+茎
+焚
+赋
+萱
+谑
+汁
+铒
+瞎
+夺
+蜗
+野
+娆
+冀
+弯
+篁
+懵
+灞
+隽
+芡
+脘
+俐
+辩
+芯
+掺
+喏
+膈
+蝈
+觐
+悚
+踹
+蔗
+熠
+鼠
+呵
+抓
+橼
+峨
+畜
+缔
+禾
+崭
+弃
+熊
+摒
+凸
+拗
+穹
+蒙
+抒
+祛
+劝
+闫
+扳
+阵
+醌
+踪
+喵
+侣
+搬
+仅
+荧
+赎
+蝾
+琦
+买
+婧
+瞄
+寓
+皎
+冻
+赝
+箩
+莫
+瞰
+郊
+笫
+姝
+筒
+枪
+遣
+煸
+袋
+舆
+痱
+涛
+母
+〇
+启
+践
+耙
+绲
+盘
+遂
+昊
+搞
+槿
+诬
+纰
+泓
+惨
+檬
+亻
+越
+C
+o
+憩
+熵
+祷
+钒
+暧
+塔
+阗
+胰
+咄
+娶
+魔
+琶
+钞
+邻
+扬
+杉
+殴
+咽
+弓
+〆
+髻
+】
+吭
+揽
+霆
+拄
+殖
+脆
+彻
+岩
+芝
+勃
+辣
+剌
+钝
+嘎
+甄
+佘
+皖
+伦
+授
+��
+憔
+挪
+皇
+庞
+稔
+芜
+踏
+溴
+兖
+卒
+擢
+饥
+鳞
+煲
+‰
+账
+颗
+叻
+斯
+捧
+鳍
+琮
+讹
+蛙
+纽
+谭
+酸
+兔
+莒
+睇
+伟
+觑
+羲
+嗜
+宜
+褐
+旎
+辛
+卦
+诘
+筋
+鎏
+溪
+挛
+熔
+阜
+晰
+鳅
+丢
+奚
+灸
+呱
+献
+陉
+黛
+鸪
+甾
+萨
+疮
+拯
+洲
+疹
+辑
+叙
+恻
+谒
+允
+柔
+烂
+氏
+逅
+漆
+拎
+惋
+扈
+湟
+纭
+啕
+掬
+擞
+哥
+忽
+涤
+鸵
+靡
+郗
+瓷
+扁
+廊
+怨
+雏
+钮
+敦
+E
+懦
+憋
+汀
+拚
+啉
+腌
+岸
+f
+痼
+瞅
+尊
+咀
+眩
+飙
+忌
+仝
+迦
+熬
+毫
+胯
+篑
+茄
+腺
+凄
+舛
+碴
+锵
+诧
+羯
+後
+漏
+汤
+宓
+仞
+蚁
+壶
+谰
+皑
+铄
+棰
+罔
+辅
+晶
+苦
+牟
+闽
+\
+烃
+饮
+聿
+丙
+蛳
+朱
+煤
+涔
+鳖
+犁
+罐
+荼
+砒
+淦
+妤
+黏
+戎
+孑
+婕
+瑾
+戢
+钵
+枣
+捋
+砥
+衩
+狙
+桠
+稣
+阎
+肃
+梏
+诫
+孪
+昶
+婊
+衫
+嗔
+侃
+塞
+蜃
+樵
+峒
+貌
+屿
+欺
+缫
+阐
+栖
+诟
+珞
+荭
+吝
+萍
+嗽
+恂
+啻
+蜴
+磬
+峋
+俸
+豫
+谎
+徊
+镍
+韬
+魇
+晴
+U
+囟
+猜
+蛮
+坐
+囿
+伴
+亭
+肝
+佗
+蝠
+妃
+胞
+滩
+榴
+氖
+垩
+苋
+砣
+扪
+馏
+姓
+轩
+厉
+夥
+侈
+禀
+垒
+岑
+赏
+钛
+辐
+痔
+披
+纸
+碳
+“
+坞
+蠓
+挤
+荥
+沅
+悔
+铧
+帼
+蒌
+蝇
+a
+p
+y
+n
+g
+哀
+浆
+瑶
+凿
+桶
+馈
+皮
+奴
+苜
+佤
+伶
+晗
+铱
+炬
+优
+弊
+氢
+恃
+甫
+攥
+端
+锌
+灰
+稹
+炝
+曙
+邋
+亥
+眶
+碾
+拉
+萝
+绔
+捷
+浍
+腋
+姑
+菖
+凌
+涞
+麽
+锢
+桨
+潢
+绎
+镰
+殆
+锑
+渝
+铬
+困
+绽
+觎
+匈
+糙
+暑
+裹
+鸟
+盔
+肽
+迷
+綦
+『
+亳
+佝
+俘
+钴
+觇
+骥
+仆
+疝
+跪
+婶
+郯
+瀹
+唉
+脖
+踞
+针
+晾
+忒
+扼
+瞩
+叛
+椒
+疟
+嗡
+邗
+肆
+跆
+玫
+忡
+捣
+咧
+唆
+艄
+蘑
+潦
+笛
+阚
+沸
+泻
+掊
+菽
+贫
+斥
+髂
+孢
+镂
+赂
+麝
+鸾
+屡
+衬
+苷
+恪
+叠
+希
+粤
+爻
+喝
+茫
+惬
+郸
+绻
+庸
+撅
+碟
+宄
+妹
+膛
+叮
+饵
+崛
+嗲
+椅
+冤
+搅
+咕
+敛
+尹
+垦
+闷
+蝉
+霎
+勰
+败
+蓑
+泸
+肤
+鹌
+幌
+焦
+浠
+鞍
+刁
+舰
+乙
+竿
+裔
+。
+茵
+函
+伊
+兄
+丨
+娜
+匍
+謇
+莪
+宥
+似
+蝽
+翳
+酪
+翠
+粑
+薇
+祢
+骏
+赠
+叫
+Q
+噤
+噻
+竖
+芗
+莠
+潭
+俊
+羿
+耜
+O
+郫
+趁
+嗪
+囚
+蹶
+芒
+洁
+笋
+鹑
+敲
+硝
+啶
+堡
+渲
+揩
+』
+携
+宿
+遒
+颍
+扭
+棱
+割
+萜
+蔸
+葵
+琴
+捂
+饰
+衙
+耿
+掠
+募
+岂
+窖
+涟
+蔺
+瘤
+柞
+瞪
+怜
+匹
+距
+楔
+炜
+哆
+秦
+缎
+幼
+茁
+绪
+痨
+恨
+楸
+娅
+瓦
+桩
+雪
+嬴
+伏
+榔
+妥
+铿
+拌
+眠
+雍
+缇
+‘
+卓
+搓
+哌
+觞
+噩
+屈
+哧
+髓
+咦
+巅
+娑
+侑
+淫
+膳
+祝
+勾
+姊
+莴
+胄
+疃
+薛
+蜷
+胛
+巷
+芙
+芋
+熙
+闰
+勿
+窃
+狱
+剩
+钏
+幢
+陟
+铛
+慧
+靴
+耍
+k
+浙
+浇
+飨
+惟
+绗
+祜
+澈
+啼
+咪
+磷
+摞
+诅
+郦
+抹
+跃
+壬
+吕
+肖
+琏
+颤
+尴
+剡
+抠
+凋
+赚
+泊
+津
+宕
+殷
+倔
+氲
+漫
+邺
+涎
+怠
+$
+垮
+荬
+遵
+俏
+叹
+噢
+饽
+蜘
+孙
+筵
+疼
+鞭
+羧
+牦
+箭
+潴
+c
+眸
+祭
+髯
+啖
+坳
+愁
+芩
+驮
+倡
+巽
+穰
+沃
+胚
+怒
+凤
+槛
+剂
+趵
+嫁
+v
+邢
+灯
+鄢
+桐
+睽
+檗
+锯
+槟
+婷
+嵋
+圻
+诗
+蕈
+颠
+遭
+痢
+芸
+怯
+馥
+竭
+锗
+徜
+恭
+遍
+籁
+剑
+嘱
+苡
+龄
+僧
+桑
+潸
+弘
+澶
+楹
+悲
+讫
+愤
+腥
+悸
+谍
+椹
+呢
+桓
+葭
+攫
+阀
+翰
+躲
+敖
+柑
+郎
+笨
+橇
+呃
+魁
+燎
+脓
+葩
+磋
+垛
+玺
+狮
+沓
+砜
+蕊
+锺
+罹
+蕉
+翱
+虐
+闾
+巫
+旦
+茱
+嬷
+枯
+鹏
+贡
+芹
+汛
+矫
+绁
+拣
+禺
+佃
+讣
+舫
+惯
+乳
+趋
+疲
+挽
+岚
+虾
+衾
+蠹
+蹂
+飓
+氦
+铖
+孩
+稞
+瑜
+壅
+掀
+勘
+妓
+畅
+髋
+W
+庐
+牲
+蓿
+榕
+练
+垣
+唱
+邸
+菲
+昆
+婺
+穿
+绡
+麒
+蚱
+掂
+愚
+泷
+涪
+漳
+妩
+娉
+榄
+讷
+觅
+旧
+藤
+煮
+呛
+柳
+腓
+叭
+庵
+烷
+阡
+罂
+蜕
+擂
+猖
+咿
+媲
+脉
+【
+沏
+貅
+黠
+熏
+哲
+烁
+坦
+酵
+兜
+×
+潇
+撒
+剽
+珩
+圹
+乾
+摸
+樟
+帽
+嗒
+襄
+魂
+轿
+憬
+锡
+〕
+喃
+皆
+咖
+隅
+脸
+残
+泮
+袂
+鹂
+珊
+囤
+捆
+咤
+误
+徨
+闹
+淙
+芊
+淋
+怆
+囗
+拨
+梳
+渤
+R
+G
+绨
+蚓
+婀
+幡
+狩
+麾
+谢
+唢
+裸
+旌
+伉
+纶
+裂
+驳
+砼
+咛
+澄
+樨
+蹈
+宙
+澍
+倍
+貔
+操
+勇
+蟠
+摈
+砧
+虬
+够
+缁
+悦
+藿
+撸
+艹
+摁
+淹
+豇
+虎
+榭
+ˉ
+吱
+d
+°
+喧
+荀
+踱
+侮
+奋
+偕
+饷
+犍
+惮
+坑
+璎
+徘
+宛
+妆
+袈
+倩
+窦
+昂
+荏
+乖
+K
+怅
+撰
+鳙
+牙
+袁
+酞
+X
+痿
+琼
+闸
+雁
+趾
+荚
+虻
+涝
+《
+杏
+韭
+偈
+烤
+绫
+鞘
+卉
+症
+遢
+蓥
+诋
+杭
+荨
+匆
+竣
+簪
+辙
+敕
+虞
+丹
+缭
+咩
+黟
+m
+淤
+瑕
+咂
+铉
+硼
+茨
+嶂
+痒
+畸
+敬
+涿
+粪
+窘
+熟
+叔
+嫔
+盾
+忱
+裘
+憾
+梵
+赡
+珙
+咯
+娘
+庙
+溯
+胺
+葱
+痪
+摊
+荷
+卞
+乒
+髦
+寐
+铭
+坩
+胗
+枷
+爆
+溟
+嚼
+羚
+砬
+轨
+惊
+挠
+罄
+竽
+菏
+氧
+浅
+楣
+盼
+枢
+炸
+阆
+杯
+谏
+噬
+淇
+渺
+俪
+秆
+墓
+泪
+跻
+砌
+痰
+垡
+渡
+耽
+釜
+讶
+鳎
+煞
+呗
+韶
+舶
+绷
+鹳
+缜
+旷
+铊
+皱
+龌
+檀
+霖
+奄
+槐
+艳
+蝶
+旋
+哝
+赶
+骞
+蚧
+腊
+盈
+丁
+`
+蜚
+矸
+蝙
+睨
+嚓
+僻
+鬼
+醴
+夜
+彝
+磊
+笔
+拔
+栀
+糕
+厦
+邰
+纫
+逭
+纤
+眦
+膊
+馍
+躇
+烯
+蘼
+冬
+诤
+暄
+骶
+哑
+瘠
+」
+臊
+丕
+愈
+咱
+螺
+擅
+跋
+搏
+硪
+谄
+笠
+淡
+嘿
+骅
+谧
+鼎
+皋
+姚
+歼
+蠢
+驼
+耳
+胬
+挝
+涯
+狗
+蒽
+孓
+犷
+凉
+芦
+箴
+铤
+孤
+嘛
+坤
+V
+茴
+朦
+挞
+尖
+橙
+诞
+搴
+碇
+洵
+浚
+帚
+蜍
+漯
+柘
+嚎
+讽
+芭
+荤
+咻
+祠
+秉
+跖
+埃
+吓
+糯
+眷
+馒
+惹
+娼
+鲑
+嫩
+讴
+轮
+瞥
+靶
+褚
+乏
+缤
+宋
+帧
+删
+驱
+碎
+扑
+俩
+俄
+偏
+涣
+竹
+噱
+皙
+佰
+渚
+唧
+斡
+#
+镉
+刀
+崎
+筐
+佣
+夭
+贰
+肴
+峙
+哔
+艿
+匐
+牺
+镛
+缘
+仡
+嫡
+劣
+枸
+堀
+梨
+簿
+鸭
+蒸
+亦
+稽
+浴
+{
+衢
+束
+槲
+j
+阁
+揍
+疥
+棋
+潋
+聪
+窜
+乓
+睛
+插
+冉
+阪
+苍
+搽
+「
+蟾
+螟
+幸
+仇
+樽
+撂
+慢
+跤
+幔
+俚
+淅
+覃
+觊
+溶
+妖
+帛
+侨
+曰
+妾
+泗
+·
+：
+瀘
+風
+Ë
+（
+）
+∶
+紅
+紗
+瑭
+雲
+頭
+鶏
+財
+許
+•
+¥
+樂
+焗
+麗
+—
+；
+滙
+東
+榮
+繪
+興
+…
+門
+業
+π
+楊
+國
+顧
+é
+盤
+寳
+Λ
+龍
+鳳
+島
+誌
+緣
+結
+銭
+萬
+勝
+祎
+璟
+優
+歡
+臨
+時
+購
+＝
+★
+藍
+昇
+鐵
+觀
+勅
+農
+聲
+畫
+兿
+術
+發
+劉
+記
+專
+耑
+園
+書
+壴
+種
+Ο
+●
+褀
+號
+銀
+匯
+敟
+锘
+葉
+橪
+廣
+進
+蒄
+鑽
+阝
+祙
+貢
+鍋
+豊
+夬
+喆
+團
+閣
+開
+燁
+賓
+館
+酡
+沔
+順
+＋
+硚
+劵
+饸
+陽
+車
+湓
+復
+萊
+氣
+軒
+華
+堃
+迮
+纟
+戶
+馬
+學
+裡
+電
+嶽
+獨
+マ
+シ
+サ
+ジ
+燘
+袪
+環
+❤
+臺
+灣
+専
+賣
+孖
+聖
+攝
+線
+▪
+α
+傢
+俬
+夢
+達
+莊
+喬
+貝
+薩
+劍
+羅
+壓
+棛
+饦
+尃
+璈
+囍
+醫
+Ｇ
+Ｉ
+Ａ
+＃
+Ｎ
+鷄
+髙
+嬰
+啓
+約
+隹
+潔
+賴
+藝
+～
+寶
+籣
+麺
+　
+嶺
+√
+義
+網
+峩
+長
+∧
+魚
+機
+構
+②
+鳯
+偉
+Ｌ
+Ｂ
+㙟
+畵
+鴿
+＇
+詩
+溝
+嚞
+屌
+藔
+佧
+玥
+蘭
+織
+１
+３
+９
+０
+７
+點
+砭
+鴨
+鋪
+銘
+廳
+弍
+‧
+創
+湯
+坶
+℃
+卩
+骝
+＆
+烜
+荘
+當
+潤
+扞
+係
+懷
+碶
+钅
+蚨
+讠
+☆
+叢
+爲
+埗
+涫
+塗
+→
+楽
+現
+鯨
+愛
+瑪
+鈺
+忄
+悶
+藥
+飾
+樓
+視
+孬
+ㆍ
+燚
+苪
+師
+①
+丼
+锽
+│
+韓
+標
+è
+兒
+閏
+匋
+張
+漢
+Ü
+髪
+會
+閑
+檔
+習
+裝
+の
+峯
+菘
+輝
+И
+雞
+釣
+億
+浐
+Ｋ
+Ｏ
+Ｒ
+８
+Ｈ
+Ｅ
+Ｐ
+Ｔ
+Ｗ
+Ｄ
+Ｓ
+Ｃ
+Ｍ
+Ｆ
+姌
+饹
+»
+晞
+廰
+ä
+嵯
+鷹
+負
+飲
+絲
+冚
+楗
+澤
+綫
+區
+❋
+←
+質
+靑
+揚
+③
+滬
+統
+産
+協
+﹑
+乸
+畐
+經
+運
+際
+洺
+岽
+為
+粵
+諾
+崋
+豐
+碁
+ɔ
+Ｖ
+２
+６
+齋
+誠
+訂
+´
+勑
+雙
+陳
+無
+í
+泩
+媄
+夌
+刂
+ｉ
+ｃ
+ｔ
+ｏ
+ｒ
+ａ
+嘢
+耄
+燴
+暃
+壽
+媽
+靈
+抻
+體
+唻
+É
+冮
+甹
+鎮
+錦
+ʌ
+蜛
+蠄
+尓
+駕
+戀
+飬
+逹
+倫
+貴
+極
+Я
+Й
+寬
+磚
+嶪
+郎
+職
+｜
+間
+ｎ
+ｄ
+剎
+伈
+課
+飛
+橋
+瘊
+№
+譜
+骓
+圗
+滘
+縣
+粿
+咅
+養
+濤
+彳
+®
+％
+Ⅱ
+啰
+㴪
+見
+矞
+薬
+糁
+邨
+鲮
+顔
+罱
+З
+選
+話
+贏
+氪
+俵
+競
+瑩
+繡
+枱
+β
+綉
+á
+獅
+爾
+™
+麵
+戋
+淩
+徳
+個
+劇
+場
+務
+簡
+寵
+ｈ
+實
+膠
+轱
+圖
+築
+嘣
+樹
+㸃
+營
+耵
+孫
+饃
+鄺
+飯
+麯
+遠
+輸
+坫
+孃
+乚
+閃
+鏢
+㎡
+題
+廠
+關
+↑
+爺
+將
+軍
+連
+篦
+覌
+參
+箸
+－
+窠
+棽
+寕
+夀
+爰
+歐
+呙
+閥
+頡
+熱
+雎
+垟
+裟
+凬
+勁
+帑
+馕
+夆
+疌
+枼
+馮
+貨
+蒤
+樸
+彧
+旸
+靜
+龢
+暢
+㐱
+鳥
+珺
+鏡
+灡
+爭
+堷
+廚
+Ó
+騰
+診
+┅
+蘇
+褔
+凱
+頂
+豕
+亞
+帥
+嘬
+⊥
+仺
+桖
+複
+饣
+絡
+穂
+顏
+棟
+納
+▏
+濟
+親
+設
+計
+攵
+埌
+烺
+ò
+頤
+燦
+蓮
+撻
+節
+講
+濱
+濃
+娽
+洳
+朿
+燈
+鈴
+護
+膚
+铔
+過
+補
+Ｚ
+Ｕ
+５
+４
+坋
+闿
+䖝
+餘
+缐
+铞
+貿
+铪
+桼
+趙
+鍊
+［
+㐂
+垚
+菓
+揸
+捲
+鐘
+滏
+𣇉
+爍
+輪
+燜
+鴻
+鮮
+動
+鹞
+鷗
+丄
+慶
+鉌
+翥
+飮
+腸
+⇋
+漁
+覺
+來
+熘
+昴
+翏
+鲱
+圧
+鄉
+萭
+頔
+爐
+嫚
+г
+貭
+類
+聯
+幛
+輕
+訓
+鑒
+夋
+锨
+芃
+珣
+䝉
+扙
+嵐
+銷
+處
+ㄱ
+語
+誘
+苝
+歸
+儀
+燒
+楿
+內
+粢
+葒
+奧
+麥
+礻
+滿
+蠔
+穵
+瞭
+態
+鱬
+榞
+硂
+鄭
+黃
+煙
+祐
+奓
+逺
+＊
+瑄
+獲
+聞
+薦
+讀
+這
+樣
+決
+問
+啟
+們
+執
+説
+轉
+單
+隨
+唘
+帶
+倉
+庫
+還
+贈
+尙
+皺
+■
+餅
+產
+○
+∈
+報
+狀
+楓
+賠
+琯
+嗮
+禮
+｀
+傳
+＞
+≤
+嗞
+Φ
+≥
+換
+咭
+∣
+↓
+曬
+ε
+応
+寫
+″
+終
+様
+純
+費
+療
+聨
+凍
+壐
+郵
+ü
+黒
+∫
+製
+塊
+調
+軽
+確
+撃
+級
+馴
+Ⅲ
+涇
+繹
+數
+碼
+證
+狒
+処
+劑
+＜
+晧
+賀
+衆
+］
+櫥
+兩
+陰
+絶
+對
+鯉
+憶
+◎
+ｐ
+ｅ
+Ｙ
+蕒
+煖
+頓
+測
+試
+鼽
+僑
+碩
+妝
+帯
+≈
+鐡
+舖
+權
+喫
+倆
+ˋ
+該
+悅
+ā
+俫
+．
+ｆ
+ｓ
+ｂ
+ｍ
+ｋ
+ｇ
+ｕ
+ｊ
+貼
+淨
+濕
+針
+適
+備
+ｌ
+／
+給
+謢
+強
+觸
+衛
+與
+⊙
+＄
+緯
+變
+⑴
+⑵
+⑶
+㎏
+殺
+∩
+幚
+─
+價
+▲
+離
+ú
+ó
+飄
+烏
+関
+閟
+﹝
+﹞
+邏
+輯
+鍵
+驗
+訣
+導
+歷
+屆
+層
+▼
+儱
+錄
+熳
+ē
+艦
+吋
+錶
+辧
+飼
+顯
+④
+禦
+販
+気
+対
+枰
+閩
+紀
+幹
+瞓
+貊
+淚
+△
+眞
+墊
+Ω
+獻
+褲
+縫
+緑
+亜
+鉅
+餠
+｛
+｝
+◆
+蘆
+薈
+█
+◇
+溫
+彈
+晳
+粧
+犸
+穩
+訊
+崬
+凖
+熥
+П
+舊
+條
+紋
+圍
+Ⅳ
+筆
+尷
+難
+雜
+錯
+綁
+識
+頰
+鎖
+艶
+□
+殁
+殼
+⑧
+├
+▕
+鵬
+ǐ
+ō
+ǒ
+糝
+綱
+▎
+μ
+盜
+饅
+醬
+籤
+蓋
+釀
+鹽
+據
+à
+ɡ
+辦
+◥
+彐
+┌
+婦
+獸
+鲩
+伱
+ī
+蒟
+蒻
+齊
+袆
+腦
+寧
+凈
+妳
+煥
+詢
+偽
+謹
+啫
+鯽
+騷
+鱸
+損
+傷
+鎻
+髮
+買
+冏
+儥
+両
+﹢
+∞
+載
+喰
+ｚ
+羙
+悵
+燙
+曉
+員
+組
+徹
+艷
+痠
+鋼
+鼙
+縮
+細
+嚒
+爯
+≠
+維
+＂
+鱻
+壇
+厍
+帰
+浥
+犇
+薡
+軎
+²
+應
+醜
+刪
+緻
+鶴
+賜
+噁
+軌
+尨
+镔
+鷺
+槗
+彌
+葚
+濛
+請
+溇
+緹
+賢
+訪
+獴
+瑅
+資
+縤
+陣
+蕟
+栢
+韻
+祼
+恁
+伢
+謝
+劃
+涑
+總
+衖
+踺
+砋
+凉
+籃
+駿
+苼
+瘋
+昽
+紡
+驊
+腎
+﹗
+響
+杋
+剛
+嚴
+禪
+歓
+槍
+傘
+檸
+檫
+炣
+勢
+鏜
+鎢
+銑
+尐
+減
+奪
+惡
+θ
+僮
+婭
+臘
+ū
+ì
+殻
+鉄
+∑
+蛲
+焼
+緖
+續
+紹
+懮

deepdoc/vision/operators.py

@@ -0,0 +1,711 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+import sys
+import six
+import cv2
+import numpy as np
+import math
+from PIL import Image
+
+
+class DecodeImage(object):
+    """ decode image """
+
+    def __init__(self,
+                 img_mode='RGB',
+                 channel_first=False,
+                 ignore_orientation=False,
+                 **kwargs):
+        self.img_mode = img_mode
+        self.channel_first = channel_first
+        self.ignore_orientation = ignore_orientation
+
+    def __call__(self, data):
+        img = data['image']
+        if six.PY2:
+            assert isinstance(img, str) and len(
+                img) > 0, "invalid input 'img' in DecodeImage"
+        else:
+            assert isinstance(img, bytes) and len(
+                img) > 0, "invalid input 'img' in DecodeImage"
+        img = np.frombuffer(img, dtype='uint8')
+        if self.ignore_orientation:
+            img = cv2.imdecode(img, cv2.IMREAD_IGNORE_ORIENTATION |
+                               cv2.IMREAD_COLOR)
+        else:
+            img = cv2.imdecode(img, 1)
+        if img is None:
+            return None
+        if self.img_mode == 'GRAY':
+            img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
+        elif self.img_mode == 'RGB':
+            assert img.shape[2] == 3, 'invalid shape of image[%s]' % (
+                img.shape)
+            img = img[:, :, ::-1]
+
+        if self.channel_first:
+            img = img.transpose((2, 0, 1))
+
+        data['image'] = img
+        return data
+
+
+class StandardizeImage(object):
+    """normalize image
+    Args:
+        mean (list): im - mean
+        std (list): im / std
+        is_scale (bool): whether need im / 255
+        norm_type (str): type in ['mean_std', 'none']
+    """
+
+    def __init__(self, mean, std, is_scale=True, norm_type='mean_std'):
+        self.mean = mean
+        self.std = std
+        self.is_scale = is_scale
+        self.norm_type = norm_type
+
+    def __call__(self, im, im_info):
+        """
+        Args:
+            im (np.ndarray): image (np.ndarray)
+            im_info (dict): info of image
+        Returns:
+            im (np.ndarray):  processed image (np.ndarray)
+            im_info (dict): info of processed image
+        """
+        im = im.astype(np.float32, copy=False)
+        if self.is_scale:
+            scale = 1.0 / 255.0
+            im *= scale
+
+        if self.norm_type == 'mean_std':
+            mean = np.array(self.mean)[np.newaxis, np.newaxis, :]
+            std = np.array(self.std)[np.newaxis, np.newaxis, :]
+            im -= mean
+            im /= std
+        return im, im_info
+
+
+class NormalizeImage(object):
+    """ normalize image such as substract mean, divide std
+    """
+
+    def __init__(self, scale=None, mean=None, std=None, order='chw', **kwargs):
+        if isinstance(scale, str):
+            scale = eval(scale)
+        self.scale = np.float32(scale if scale is not None else 1.0 / 255.0)
+        mean = mean if mean is not None else [0.485, 0.456, 0.406]
+        std = std if std is not None else [0.229, 0.224, 0.225]
+
+        shape = (3, 1, 1) if order == 'chw' else (1, 1, 3)
+        self.mean = np.array(mean).reshape(shape).astype('float32')
+        self.std = np.array(std).reshape(shape).astype('float32')
+
+    def __call__(self, data):
+        img = data['image']
+        from PIL import Image
+        if isinstance(img, Image.Image):
+            img = np.array(img)
+        assert isinstance(img,
+                          np.ndarray), "invalid input 'img' in NormalizeImage"
+        data['image'] = (
+            img.astype('float32') * self.scale - self.mean) / self.std
+        return data
+
+
+class ToCHWImage(object):
+    """ convert hwc image to chw image
+    """
+
+    def __init__(self, **kwargs):
+        pass
+
+    def __call__(self, data):
+        img = data['image']
+        from PIL import Image
+        if isinstance(img, Image.Image):
+            img = np.array(img)
+        data['image'] = img.transpose((2, 0, 1))
+        return data
+
+
+class Fasttext(object):
+    def __init__(self, path="None", **kwargs):
+        import fasttext
+        self.fast_model = fasttext.load_model(path)
+
+    def __call__(self, data):
+        label = data['label']
+        fast_label = self.fast_model[label]
+        data['fast_label'] = fast_label
+        return data
+
+
+class KeepKeys(object):
+    def __init__(self, keep_keys, **kwargs):
+        self.keep_keys = keep_keys
+
+    def __call__(self, data):
+        data_list = []
+        for key in self.keep_keys:
+            data_list.append(data[key])
+        return data_list
+
+
+class Pad(object):
+    def __init__(self, size=None, size_div=32, **kwargs):
+        if size is not None and not isinstance(size, (int, list, tuple)):
+            raise TypeError("Type of target_size is invalid. Now is {}".format(
+                type(size)))
+        if isinstance(size, int):
+            size = [size, size]
+        self.size = size
+        self.size_div = size_div
+
+    def __call__(self, data):
+
+        img = data['image']
+        img_h, img_w = img.shape[0], img.shape[1]
+        if self.size:
+            resize_h2, resize_w2 = self.size
+            assert (
+                img_h < resize_h2 and img_w < resize_w2
+            ), '(h, w) of target size should be greater than (img_h, img_w)'
+        else:
+            resize_h2 = max(
+                int(math.ceil(img.shape[0] / self.size_div) * self.size_div),
+                self.size_div)
+            resize_w2 = max(
+                int(math.ceil(img.shape[1] / self.size_div) * self.size_div),
+                self.size_div)
+        img = cv2.copyMakeBorder(
+            img,
+            0,
+            resize_h2 - img_h,
+            0,
+            resize_w2 - img_w,
+            cv2.BORDER_CONSTANT,
+            value=0)
+        data['image'] = img
+        return data
+
+
+class LinearResize(object):
+    """resize image by target_size and max_size
+    Args:
+        target_size (int): the target size of image
+        keep_ratio (bool): whether keep_ratio or not, default true
+        interp (int): method of resize
+    """
+
+    def __init__(self, target_size, keep_ratio=True, interp=cv2.INTER_LINEAR):
+        if isinstance(target_size, int):
+            target_size = [target_size, target_size]
+        self.target_size = target_size
+        self.keep_ratio = keep_ratio
+        self.interp = interp
+
+    def __call__(self, im, im_info):
+        """
+        Args:
+            im (np.ndarray): image (np.ndarray)
+            im_info (dict): info of image
+        Returns:
+            im (np.ndarray):  processed image (np.ndarray)
+            im_info (dict): info of processed image
+        """
+        assert len(self.target_size) == 2
+        assert self.target_size[0] > 0 and self.target_size[1] > 0
+        im_channel = im.shape[2]
+        im_scale_y, im_scale_x = self.generate_scale(im)
+        im = cv2.resize(
+            im,
+            None,
+            None,
+            fx=im_scale_x,
+            fy=im_scale_y,
+            interpolation=self.interp)
+        im_info['im_shape'] = np.array(im.shape[:2]).astype('float32')
+        im_info['scale_factor'] = np.array(
+            [im_scale_y, im_scale_x]).astype('float32')
+        return im, im_info
+
+    def generate_scale(self, im):
+        """
+        Args:
+            im (np.ndarray): image (np.ndarray)
+        Returns:
+            im_scale_x: the resize ratio of X
+            im_scale_y: the resize ratio of Y
+        """
+        origin_shape = im.shape[:2]
+        im_c = im.shape[2]
+        if self.keep_ratio:
+            im_size_min = np.min(origin_shape)
+            im_size_max = np.max(origin_shape)
+            target_size_min = np.min(self.target_size)
+            target_size_max = np.max(self.target_size)
+            im_scale = float(target_size_min) / float(im_size_min)
+            if np.round(im_scale * im_size_max) > target_size_max:
+                im_scale = float(target_size_max) / float(im_size_max)
+            im_scale_x = im_scale
+            im_scale_y = im_scale
+        else:
+            resize_h, resize_w = self.target_size
+            im_scale_y = resize_h / float(origin_shape[0])
+            im_scale_x = resize_w / float(origin_shape[1])
+        return im_scale_y, im_scale_x
+
+
+class Resize(object):
+    def __init__(self, size=(640, 640), **kwargs):
+        self.size = size
+
+    def resize_image(self, img):
+        resize_h, resize_w = self.size
+        ori_h, ori_w = img.shape[:2]  # (h, w, c)
+        ratio_h = float(resize_h) / ori_h
+        ratio_w = float(resize_w) / ori_w
+        img = cv2.resize(img, (int(resize_w), int(resize_h)))
+        return img, [ratio_h, ratio_w]
+
+    def __call__(self, data):
+        img = data['image']
+        if 'polys' in data:
+            text_polys = data['polys']
+
+        img_resize, [ratio_h, ratio_w] = self.resize_image(img)
+        if 'polys' in data:
+            new_boxes = []
+            for box in text_polys:
+                new_box = []
+                for cord in box:
+                    new_box.append([cord[0] * ratio_w, cord[1] * ratio_h])
+                new_boxes.append(new_box)
+            data['polys'] = np.array(new_boxes, dtype=np.float32)
+        data['image'] = img_resize
+        return data
+
+
+class DetResizeForTest(object):
+    def __init__(self, **kwargs):
+        super(DetResizeForTest, self).__init__()
+        self.resize_type = 0
+        self.keep_ratio = False
+        if 'image_shape' in kwargs:
+            self.image_shape = kwargs['image_shape']
+            self.resize_type = 1
+            if 'keep_ratio' in kwargs:
+                self.keep_ratio = kwargs['keep_ratio']
+        elif 'limit_side_len' in kwargs:
+            self.limit_side_len = kwargs['limit_side_len']
+            self.limit_type = kwargs.get('limit_type', 'min')
+        elif 'resize_long' in kwargs:
+            self.resize_type = 2
+            self.resize_long = kwargs.get('resize_long', 960)
+        else:
+            self.limit_side_len = 736
+            self.limit_type = 'min'
+
+    def __call__(self, data):
+        img = data['image']
+        src_h, src_w, _ = img.shape
+        if sum([src_h, src_w]) < 64:
+            img = self.image_padding(img)
+
+        if self.resize_type == 0:
+            # img, shape = self.resize_image_type0(img)
+            img, [ratio_h, ratio_w] = self.resize_image_type0(img)
+        elif self.resize_type == 2:
+            img, [ratio_h, ratio_w] = self.resize_image_type2(img)
+        else:
+            # img, shape = self.resize_image_type1(img)
+            img, [ratio_h, ratio_w] = self.resize_image_type1(img)
+        data['image'] = img
+        data['shape'] = np.array([src_h, src_w, ratio_h, ratio_w])
+        return data
+
+    def image_padding(self, im, value=0):
+        h, w, c = im.shape
+        im_pad = np.zeros((max(32, h), max(32, w), c), np.uint8) + value
+        im_pad[:h, :w, :] = im
+        return im_pad
+
+    def resize_image_type1(self, img):
+        resize_h, resize_w = self.image_shape
+        ori_h, ori_w = img.shape[:2]  # (h, w, c)
+        if self.keep_ratio is True:
+            resize_w = ori_w * resize_h / ori_h
+            N = math.ceil(resize_w / 32)
+            resize_w = N * 32
+        ratio_h = float(resize_h) / ori_h
+        ratio_w = float(resize_w) / ori_w
+        img = cv2.resize(img, (int(resize_w), int(resize_h)))
+        # return img, np.array([ori_h, ori_w])
+        return img, [ratio_h, ratio_w]
+
+    def resize_image_type0(self, img):
+        """
+        resize image to a size multiple of 32 which is required by the network
+        args:
+            img(array): array with shape [h, w, c]
+        return(tuple):
+            img, (ratio_h, ratio_w)
+        """
+        limit_side_len = self.limit_side_len
+        h, w, c = img.shape
+
+        # limit the max side
+        if self.limit_type == 'max':
+            if max(h, w) > limit_side_len:
+                if h > w:
+                    ratio = float(limit_side_len) / h
+                else:
+                    ratio = float(limit_side_len) / w
+            else:
+                ratio = 1.
+        elif self.limit_type == 'min':
+            if min(h, w) < limit_side_len:
+                if h < w:
+                    ratio = float(limit_side_len) / h
+                else:
+                    ratio = float(limit_side_len) / w
+            else:
+                ratio = 1.
+        elif self.limit_type == 'resize_long':
+            ratio = float(limit_side_len) / max(h, w)
+        else:
+            raise Exception('not support limit type, image ')
+        resize_h = int(h * ratio)
+        resize_w = int(w * ratio)
+
+        resize_h = max(int(round(resize_h / 32) * 32), 32)
+        resize_w = max(int(round(resize_w / 32) * 32), 32)
+
+        try:
+            if int(resize_w) <= 0 or int(resize_h) <= 0:
+                return None, (None, None)
+            img = cv2.resize(img, (int(resize_w), int(resize_h)))
+        except BaseException:
+            print(img.shape, resize_w, resize_h)
+            sys.exit(0)
+        ratio_h = resize_h / float(h)
+        ratio_w = resize_w / float(w)
+        return img, [ratio_h, ratio_w]
+
+    def resize_image_type2(self, img):
+        h, w, _ = img.shape
+
+        resize_w = w
+        resize_h = h
+
+        if resize_h > resize_w:
+            ratio = float(self.resize_long) / resize_h
+        else:
+            ratio = float(self.resize_long) / resize_w
+
+        resize_h = int(resize_h * ratio)
+        resize_w = int(resize_w * ratio)
+
+        max_stride = 128
+        resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+        resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+        img = cv2.resize(img, (int(resize_w), int(resize_h)))
+        ratio_h = resize_h / float(h)
+        ratio_w = resize_w / float(w)
+
+        return img, [ratio_h, ratio_w]
+
+
+class E2EResizeForTest(object):
+    def __init__(self, **kwargs):
+        super(E2EResizeForTest, self).__init__()
+        self.max_side_len = kwargs['max_side_len']
+        self.valid_set = kwargs['valid_set']
+
+    def __call__(self, data):
+        img = data['image']
+        src_h, src_w, _ = img.shape
+        if self.valid_set == 'totaltext':
+            im_resized, [ratio_h, ratio_w] = self.resize_image_for_totaltext(
+                img, max_side_len=self.max_side_len)
+        else:
+            im_resized, (ratio_h, ratio_w) = self.resize_image(
+                img, max_side_len=self.max_side_len)
+        data['image'] = im_resized
+        data['shape'] = np.array([src_h, src_w, ratio_h, ratio_w])
+        return data
+
+    def resize_image_for_totaltext(self, im, max_side_len=512):
+
+        h, w, _ = im.shape
+        resize_w = w
+        resize_h = h
+        ratio = 1.25
+        if h * ratio > max_side_len:
+            ratio = float(max_side_len) / resize_h
+        resize_h = int(resize_h * ratio)
+        resize_w = int(resize_w * ratio)
+
+        max_stride = 128
+        resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+        resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+        im = cv2.resize(im, (int(resize_w), int(resize_h)))
+        ratio_h = resize_h / float(h)
+        ratio_w = resize_w / float(w)
+        return im, (ratio_h, ratio_w)
+
+    def resize_image(self, im, max_side_len=512):
+        """
+        resize image to a size multiple of max_stride which is required by the network
+        :param im: the resized image
+        :param max_side_len: limit of max image size to avoid out of memory in gpu
+        :return: the resized image and the resize ratio
+        """
+        h, w, _ = im.shape
+
+        resize_w = w
+        resize_h = h
+
+        # Fix the longer side
+        if resize_h > resize_w:
+            ratio = float(max_side_len) / resize_h
+        else:
+            ratio = float(max_side_len) / resize_w
+
+        resize_h = int(resize_h * ratio)
+        resize_w = int(resize_w * ratio)
+
+        max_stride = 128
+        resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+        resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+        im = cv2.resize(im, (int(resize_w), int(resize_h)))
+        ratio_h = resize_h / float(h)
+        ratio_w = resize_w / float(w)
+
+        return im, (ratio_h, ratio_w)
+
+
+class KieResize(object):
+    def __init__(self, **kwargs):
+        super(KieResize, self).__init__()
+        self.max_side, self.min_side = kwargs['img_scale'][0], kwargs[
+            'img_scale'][1]
+
+    def __call__(self, data):
+        img = data['image']
+        points = data['points']
+        src_h, src_w, _ = img.shape
+        im_resized, scale_factor, [ratio_h, ratio_w
+                                   ], [new_h, new_w] = self.resize_image(img)
+        resize_points = self.resize_boxes(img, points, scale_factor)
+        data['ori_image'] = img
+        data['ori_boxes'] = points
+        data['points'] = resize_points
+        data['image'] = im_resized
+        data['shape'] = np.array([new_h, new_w])
+        return data
+
+    def resize_image(self, img):
+        norm_img = np.zeros([1024, 1024, 3], dtype='float32')
+        scale = [512, 1024]
+        h, w = img.shape[:2]
+        max_long_edge = max(scale)
+        max_short_edge = min(scale)
+        scale_factor = min(max_long_edge / max(h, w),
+                           max_short_edge / min(h, w))
+        resize_w, resize_h = int(w * float(scale_factor) + 0.5), int(h * float(
+            scale_factor) + 0.5)
+        max_stride = 32
+        resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+        resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+        im = cv2.resize(img, (resize_w, resize_h))
+        new_h, new_w = im.shape[:2]
+        w_scale = new_w / w
+        h_scale = new_h / h
+        scale_factor = np.array(
+            [w_scale, h_scale, w_scale, h_scale], dtype=np.float32)
+        norm_img[:new_h, :new_w, :] = im
+        return norm_img, scale_factor, [h_scale, w_scale], [new_h, new_w]
+
+    def resize_boxes(self, im, points, scale_factor):
+        points = points * scale_factor
+        img_shape = im.shape[:2]
+        points[:, 0::2] = np.clip(points[:, 0::2], 0, img_shape[1])
+        points[:, 1::2] = np.clip(points[:, 1::2], 0, img_shape[0])
+        return points
+
+
+class SRResize(object):
+    def __init__(self,
+                 imgH=32,
+                 imgW=128,
+                 down_sample_scale=4,
+                 keep_ratio=False,
+                 min_ratio=1,
+                 mask=False,
+                 infer_mode=False,
+                 **kwargs):
+        self.imgH = imgH
+        self.imgW = imgW
+        self.keep_ratio = keep_ratio
+        self.min_ratio = min_ratio
+        self.down_sample_scale = down_sample_scale
+        self.mask = mask
+        self.infer_mode = infer_mode
+
+    def __call__(self, data):
+        imgH = self.imgH
+        imgW = self.imgW
+        images_lr = data["image_lr"]
+        transform2 = ResizeNormalize(
+            (imgW // self.down_sample_scale, imgH // self.down_sample_scale))
+        images_lr = transform2(images_lr)
+        data["img_lr"] = images_lr
+        if self.infer_mode:
+            return data
+
+        images_HR = data["image_hr"]
+        label_strs = data["label"]
+        transform = ResizeNormalize((imgW, imgH))
+        images_HR = transform(images_HR)
+        data["img_hr"] = images_HR
+        return data
+
+
+class ResizeNormalize(object):
+    def __init__(self, size, interpolation=Image.BICUBIC):
+        self.size = size
+        self.interpolation = interpolation
+
+    def __call__(self, img):
+        img = img.resize(self.size, self.interpolation)
+        img_numpy = np.array(img).astype("float32")
+        img_numpy = img_numpy.transpose((2, 0, 1)) / 255
+        return img_numpy
+
+
+class GrayImageChannelFormat(object):
+    """
+    format gray scale image's channel: (3,h,w) -> (1,h,w)
+    Args:
+        inverse: inverse gray image
+    """
+
+    def __init__(self, inverse=False, **kwargs):
+        self.inverse = inverse
+
+    def __call__(self, data):
+        img = data['image']
+        img_single_channel = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+        img_expanded = np.expand_dims(img_single_channel, 0)
+
+        if self.inverse:
+            data['image'] = np.abs(img_expanded - 1)
+        else:
+            data['image'] = img_expanded
+
+        data['src_image'] = img
+        return data
+
+
+class Permute(object):
+    """permute image
+    Args:
+        to_bgr (bool): whether convert RGB to BGR
+        channel_first (bool): whether convert HWC to CHW
+    """
+
+    def __init__(self, ):
+        super(Permute, self).__init__()
+
+    def __call__(self, im, im_info):
+        """
+        Args:
+            im (np.ndarray): image (np.ndarray)
+            im_info (dict): info of image
+        Returns:
+            im (np.ndarray):  processed image (np.ndarray)
+            im_info (dict): info of processed image
+        """
+        im = im.transpose((2, 0, 1)).copy()
+        return im, im_info
+
+
+class PadStride(object):
+    """ padding image for model with FPN, instead PadBatch(pad_to_stride) in original config
+    Args:
+        stride (bool): model with FPN need image shape % stride == 0
+    """
+
+    def __init__(self, stride=0):
+        self.coarsest_stride = stride
+
+    def __call__(self, im, im_info):
+        """
+        Args:
+            im (np.ndarray): image (np.ndarray)
+            im_info (dict): info of image
+        Returns:
+            im (np.ndarray):  processed image (np.ndarray)
+            im_info (dict): info of processed image
+        """
+        coarsest_stride = self.coarsest_stride
+        if coarsest_stride <= 0:
+            return im, im_info
+        im_c, im_h, im_w = im.shape
+        pad_h = int(np.ceil(float(im_h) / coarsest_stride) * coarsest_stride)
+        pad_w = int(np.ceil(float(im_w) / coarsest_stride) * coarsest_stride)
+        padding_im = np.zeros((im_c, pad_h, pad_w), dtype=np.float32)
+        padding_im[:, :im_h, :im_w] = im
+        return padding_im, im_info
+
+
+def decode_image(im_file, im_info):
+    """read rgb image
+    Args:
+        im_file (str|np.ndarray): input can be image path or np.ndarray
+        im_info (dict): info of image
+    Returns:
+        im (np.ndarray):  processed image (np.ndarray)
+        im_info (dict): info of processed image
+    """
+    if isinstance(im_file, str):
+        with open(im_file, 'rb') as f:
+            im_read = f.read()
+        data = np.frombuffer(im_read, dtype='uint8')
+        im = cv2.imdecode(data, 1)  # BGR mode, but need RGB mode
+        im = cv2.cvtColor(im, cv2.COLOR_BGR2RGB)
+    else:
+        im = im_file
+    im_info['im_shape'] = np.array(im.shape[:2], dtype=np.float32)
+    im_info['scale_factor'] = np.array([1., 1.], dtype=np.float32)
+    return im, im_info
+
+
+def preprocess(im, preprocess_ops):
+    # process image by preprocess_ops
+    im_info = {
+        'scale_factor': np.array(
+            [1., 1.], dtype=np.float32),
+        'im_shape': None,
+    }
+    im, im_info = decode_image(im, im_info)
+    for operator in preprocess_ops:
+        im, im_info = operator(im, im_info)
+    return im, im_info

deepdoc/vision/postprocess.py

@@ -0,0 +1,353 @@
+import copy
+
+import numpy as np
+import cv2
+from shapely.geometry import Polygon
+import pyclipper
+
+
+def build_post_process(config, global_config=None):
+    support_dict = ['DBPostProcess', 'CTCLabelDecode']
+
+    config = copy.deepcopy(config)
+    module_name = config.pop('name')
+    if module_name == "None":
+        return
+    if global_config is not None:
+        config.update(global_config)
+    assert module_name in support_dict, Exception(
+        'post process only support {}'.format(support_dict))
+    module_class = eval(module_name)(**config)
+    return module_class
+
+
+class DBPostProcess(object):
+    """
+    The post process for Differentiable Binarization (DB).
+    """
+
+    def __init__(self,
+                 thresh=0.3,
+                 box_thresh=0.7,
+                 max_candidates=1000,
+                 unclip_ratio=2.0,
+                 use_dilation=False,
+                 score_mode="fast",
+                 box_type='quad',
+                 **kwargs):
+        self.thresh = thresh
+        self.box_thresh = box_thresh
+        self.max_candidates = max_candidates
+        self.unclip_ratio = unclip_ratio
+        self.min_size = 3
+        self.score_mode = score_mode
+        self.box_type = box_type
+        assert score_mode in [
+            "slow", "fast"
+        ], "Score mode must be in [slow, fast] but got: {}".format(score_mode)
+
+        self.dilation_kernel = None if not use_dilation else np.array(
+            [[1, 1], [1, 1]])
+
+    def polygons_from_bitmap(self, pred, _bitmap, dest_width, dest_height):
+        '''
+        _bitmap: single map with shape (1, H, W),
+            whose values are binarized as {0, 1}
+        '''
+
+        bitmap = _bitmap
+        height, width = bitmap.shape
+
+        boxes = []
+        scores = []
+
+        contours, _ = cv2.findContours((bitmap * 255).astype(np.uint8),
+                                       cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
+
+        for contour in contours[:self.max_candidates]:
+            epsilon = 0.002 * cv2.arcLength(contour, True)
+            approx = cv2.approxPolyDP(contour, epsilon, True)
+            points = approx.reshape((-1, 2))
+            if points.shape[0] < 4:
+                continue
+
+            score = self.box_score_fast(pred, points.reshape(-1, 2))
+            if self.box_thresh > score:
+                continue
+
+            if points.shape[0] > 2:
+                box = self.unclip(points, self.unclip_ratio)
+                if len(box) > 1:
+                    continue
+            else:
+                continue
+            box = box.reshape(-1, 2)
+
+            _, sside = self.get_mini_boxes(box.reshape((-1, 1, 2)))
+            if sside < self.min_size + 2:
+                continue
+
+            box = np.array(box)
+            box[:, 0] = np.clip(
+                np.round(box[:, 0] / width * dest_width), 0, dest_width)
+            box[:, 1] = np.clip(
+                np.round(box[:, 1] / height * dest_height), 0, dest_height)
+            boxes.append(box.tolist())
+            scores.append(score)
+        return boxes, scores
+
+    def boxes_from_bitmap(self, pred, _bitmap, dest_width, dest_height):
+        '''
+        _bitmap: single map with shape (1, H, W),
+                whose values are binarized as {0, 1}
+        '''
+
+        bitmap = _bitmap
+        height, width = bitmap.shape
+
+        outs = cv2.findContours((bitmap * 255).astype(np.uint8), cv2.RETR_LIST,
+                                cv2.CHAIN_APPROX_SIMPLE)
+        if len(outs) == 3:
+            img, contours, _ = outs[0], outs[1], outs[2]
+        elif len(outs) == 2:
+            contours, _ = outs[0], outs[1]
+
+        num_contours = min(len(contours), self.max_candidates)
+
+        boxes = []
+        scores = []
+        for index in range(num_contours):
+            contour = contours[index]
+            points, sside = self.get_mini_boxes(contour)
+            if sside < self.min_size:
+                continue
+            points = np.array(points)
+            if self.score_mode == "fast":
+                score = self.box_score_fast(pred, points.reshape(-1, 2))
+            else:
+                score = self.box_score_slow(pred, contour)
+            if self.box_thresh > score:
+                continue
+
+            box = self.unclip(points, self.unclip_ratio).reshape(-1, 1, 2)
+            box, sside = self.get_mini_boxes(box)
+            if sside < self.min_size + 2:
+                continue
+            box = np.array(box)
+
+            box[:, 0] = np.clip(
+                np.round(box[:, 0] / width * dest_width), 0, dest_width)
+            box[:, 1] = np.clip(
+                np.round(box[:, 1] / height * dest_height), 0, dest_height)
+            boxes.append(box.astype("int32"))
+            scores.append(score)
+        return np.array(boxes, dtype="int32"), scores
+
+    def unclip(self, box, unclip_ratio):
+        poly = Polygon(box)
+        distance = poly.area * unclip_ratio / poly.length
+        offset = pyclipper.PyclipperOffset()
+        offset.AddPath(box, pyclipper.JT_ROUND, pyclipper.ET_CLOSEDPOLYGON)
+        expanded = np.array(offset.Execute(distance))
+        return expanded
+
+    def get_mini_boxes(self, contour):
+        bounding_box = cv2.minAreaRect(contour)
+        points = sorted(list(cv2.boxPoints(bounding_box)), key=lambda x: x[0])
+
+        index_1, index_2, index_3, index_4 = 0, 1, 2, 3
+        if points[1][1] > points[0][1]:
+            index_1 = 0
+            index_4 = 1
+        else:
+            index_1 = 1
+            index_4 = 0
+        if points[3][1] > points[2][1]:
+            index_2 = 2
+            index_3 = 3
+        else:
+            index_2 = 3
+            index_3 = 2
+
+        box = [
+            points[index_1], points[index_2], points[index_3], points[index_4]
+        ]
+        return box, min(bounding_box[1])
+
+    def box_score_fast(self, bitmap, _box):
+        '''
+        box_score_fast: use bbox mean score as the mean score
+        '''
+        h, w = bitmap.shape[:2]
+        box = _box.copy()
+        xmin = np.clip(np.floor(box[:, 0].min()).astype("int32"), 0, w - 1)
+        xmax = np.clip(np.ceil(box[:, 0].max()).astype("int32"), 0, w - 1)
+        ymin = np.clip(np.floor(box[:, 1].min()).astype("int32"), 0, h - 1)
+        ymax = np.clip(np.ceil(box[:, 1].max()).astype("int32"), 0, h - 1)
+
+        mask = np.zeros((ymax - ymin + 1, xmax - xmin + 1), dtype=np.uint8)
+        box[:, 0] = box[:, 0] - xmin
+        box[:, 1] = box[:, 1] - ymin
+        cv2.fillPoly(mask, box.reshape(1, -1, 2).astype("int32"), 1)
+        return cv2.mean(bitmap[ymin:ymax + 1, xmin:xmax + 1], mask)[0]
+
+    def box_score_slow(self, bitmap, contour):
+        '''
+        box_score_slow: use polyon mean score as the mean score
+        '''
+        h, w = bitmap.shape[:2]
+        contour = contour.copy()
+        contour = np.reshape(contour, (-1, 2))
+
+        xmin = np.clip(np.min(contour[:, 0]), 0, w - 1)
+        xmax = np.clip(np.max(contour[:, 0]), 0, w - 1)
+        ymin = np.clip(np.min(contour[:, 1]), 0, h - 1)
+        ymax = np.clip(np.max(contour[:, 1]), 0, h - 1)
+
+        mask = np.zeros((ymax - ymin + 1, xmax - xmin + 1), dtype=np.uint8)
+
+        contour[:, 0] = contour[:, 0] - xmin
+        contour[:, 1] = contour[:, 1] - ymin
+
+        cv2.fillPoly(mask, contour.reshape(1, -1, 2).astype("int32"), 1)
+        return cv2.mean(bitmap[ymin:ymax + 1, xmin:xmax + 1], mask)[0]
+
+    def __call__(self, outs_dict, shape_list):
+        pred = outs_dict['maps']
+        if not isinstance(pred, np.ndarray):
+            pred = pred.numpy()
+        pred = pred[:, 0, :, :]
+        segmentation = pred > self.thresh
+
+        boxes_batch = []
+        for batch_index in range(pred.shape[0]):
+            src_h, src_w, ratio_h, ratio_w = shape_list[batch_index]
+            if self.dilation_kernel is not None:
+                mask = cv2.dilate(
+                    np.array(segmentation[batch_index]).astype(np.uint8),
+                    self.dilation_kernel)
+            else:
+                mask = segmentation[batch_index]
+            if self.box_type == 'poly':
+                boxes, scores = self.polygons_from_bitmap(pred[batch_index],
+                                                          mask, src_w, src_h)
+            elif self.box_type == 'quad':
+                boxes, scores = self.boxes_from_bitmap(pred[batch_index], mask,
+                                                       src_w, src_h)
+            else:
+                raise ValueError(
+                    "box_type can only be one of ['quad', 'poly']")
+
+            boxes_batch.append({'points': boxes})
+        return boxes_batch
+
+
+class BaseRecLabelDecode(object):
+    """ Convert between text-label and text-index """
+
+    def __init__(self, character_dict_path=None, use_space_char=False):
+        self.beg_str = "sos"
+        self.end_str = "eos"
+        self.reverse = False
+        self.character_str = []
+
+        if character_dict_path is None:
+            self.character_str = "0123456789abcdefghijklmnopqrstuvwxyz"
+            dict_character = list(self.character_str)
+        else:
+            with open(character_dict_path, "rb") as fin:
+                lines = fin.readlines()
+                for line in lines:
+                    line = line.decode('utf-8').strip("\n").strip("\r\n")
+                    self.character_str.append(line)
+            if use_space_char:
+                self.character_str.append(" ")
+            dict_character = list(self.character_str)
+            if 'arabic' in character_dict_path:
+                self.reverse = True
+
+        dict_character = self.add_special_char(dict_character)
+        self.dict = {}
+        for i, char in enumerate(dict_character):
+            self.dict[char] = i
+        self.character = dict_character
+
+    def pred_reverse(self, pred):
+        pred_re = []
+        c_current = ''
+        for c in pred:
+            if not bool(re.search('[a-zA-Z0-9 :*./%+-]', c)):
+                if c_current != '':
+                    pred_re.append(c_current)
+                pred_re.append(c)
+                c_current = ''
+            else:
+                c_current += c
+        if c_current != '':
+            pred_re.append(c_current)
+
+        return ''.join(pred_re[::-1])
+
+    def add_special_char(self, dict_character):
+        return dict_character
+
+    def decode(self, text_index, text_prob=None, is_remove_duplicate=False):
+        """ convert text-index into text-label. """
+        result_list = []
+        ignored_tokens = self.get_ignored_tokens()
+        batch_size = len(text_index)
+        for batch_idx in range(batch_size):
+            selection = np.ones(len(text_index[batch_idx]), dtype=bool)
+            if is_remove_duplicate:
+                selection[1:] = text_index[batch_idx][1:] != text_index[
+                    batch_idx][:-1]
+            for ignored_token in ignored_tokens:
+                selection &= text_index[batch_idx] != ignored_token
+
+            char_list = [
+                self.character[text_id]
+                for text_id in text_index[batch_idx][selection]
+            ]
+            if text_prob is not None:
+                conf_list = text_prob[batch_idx][selection]
+            else:
+                conf_list = [1] * len(selection)
+            if len(conf_list) == 0:
+                conf_list = [0]
+
+            text = ''.join(char_list)
+
+            if self.reverse:  # for arabic rec
+                text = self.pred_reverse(text)
+
+            result_list.append((text, np.mean(conf_list).tolist()))
+        return result_list
+
+    def get_ignored_tokens(self):
+        return [0]  # for ctc blank
+
+
+class CTCLabelDecode(BaseRecLabelDecode):
+    """ Convert between text-label and text-index """
+
+    def __init__(self, character_dict_path=None, use_space_char=False,
+                 **kwargs):
+        super(CTCLabelDecode, self).__init__(character_dict_path,
+                                             use_space_char)
+
+    def __call__(self, preds, label=None, *args, **kwargs):
+        if isinstance(preds, tuple) or isinstance(preds, list):
+            preds = preds[-1]
+        if not isinstance(preds, np.ndarray):
+            preds = preds.numpy()
+        preds_idx = preds.argmax(axis=2)
+        preds_prob = preds.max(axis=2)
+        text = self.decode(preds_idx, preds_prob, is_remove_duplicate=True)
+        if label is None:
+            return text
+        label = self.decode(label)
+        return text, label
+
+    def add_special_char(self, dict_character):
+        dict_character = ['blank'] + dict_character
+        return dict_character

deepdoc/vision/ragFlow.py

@@ -0,0 +1,313 @@
+import copy
+import time
+import os
+
+from huggingface_hub import snapshot_download
+
+from .operators import *
+import numpy as np
+import onnxruntime as ort
+import logging
+from .postprocess import build_post_process
+
+from typing import List
+
+def get_deepdoc_directory():
+    PROJECT_BASE = os.path.abspath(
+            os.path.join(
+                os.path.dirname(os.path.realpath(__file__)),
+                os.pardir
+            )
+    )
+    return PROJECT_BASE
+def transform(data, ops=None):
+    """ transform """
+    if ops is None:
+        ops = []
+    for op in ops:
+        data = op(data)
+        if data is None:
+            return None
+    return data
+
+
+def create_operators(op_param_list, global_config=None):
+    """
+    create operators based on the config
+
+    Args:
+        params(list): a dict list, used to create some operators
+    """
+    assert isinstance(
+        op_param_list, list), ('operator config should be a list')
+    ops = []
+    for operator in op_param_list:
+        assert isinstance(operator,
+                          dict) and len(operator) == 1, "yaml format error"
+        op_name = list(operator)[0]
+        param = {} if operator[op_name] is None else operator[op_name]
+        if global_config is not None:
+            param.update(global_config)
+        op = eval(op_name)(**param)
+        ops.append(op)
+    return ops
+
+
+def load_model(model_dir, nm):
+    model_file_path = os.path.join(model_dir, nm + ".onnx")
+    if not os.path.exists(model_file_path):
+        raise ValueError("not find model file path {}".format(
+            model_file_path))
+
+    options = ort.SessionOptions()
+    options.enable_cpu_mem_arena = False
+    options.execution_mode = ort.ExecutionMode.ORT_SEQUENTIAL
+    options.intra_op_num_threads = 2
+    options.inter_op_num_threads = 2
+    if False and ort.get_device() == "GPU":
+        sess = ort.InferenceSession(
+            model_file_path,
+            options=options,
+            providers=['CUDAExecutionProvider'])
+    else:
+        sess = ort.InferenceSession(
+            model_file_path,
+            options=options,
+            providers=['CPUExecutionProvider'])
+        print(model_file_path)
+        print(sess.get_modelmeta().description)
+    return sess, sess.get_inputs()[0]
+
+
+class RagFlowTextDetector:
+    """
+    The  class depends on TextDetector to perform its primary function of detecting text and retrieving bounding boxes.
+    """
+    def __init__(self, model_dir):
+        pre_process_list = [{
+            'DetResizeForTest': {
+                'limit_side_len': 960,
+                'limit_type': "max",
+            }
+        }, {
+            'NormalizeImage': {
+                'std': [0.229, 0.224, 0.225],
+                'mean': [0.485, 0.456, 0.406],
+                'scale': '1./255.',
+                'order': 'hwc'
+            }
+        }, {
+            'ToCHWImage': None
+        }, {
+            'KeepKeys': {
+                'keep_keys': ['image', 'shape']
+            }
+        }]
+        postprocess_params = {"name": "DBPostProcess", "thresh": 0.3, "box_thresh": 0.5, "max_candidates": 1000,
+                              "unclip_ratio": 1.5, "use_dilation": False, "score_mode": "fast", "box_type": "quad"}
+
+        self.postprocess_op = build_post_process(postprocess_params)
+        self.predictor, self.input_tensor = load_model(model_dir, 'det')
+
+        img_h, img_w = self.input_tensor.shape[2:]
+        if isinstance(img_h, str) or isinstance(img_w, str):
+            pass
+        elif img_h is not None and img_w is not None and img_h > 0 and img_w > 0:
+            pre_process_list[0] = {
+                'DetResizeForTest': {
+                    'image_shape': [img_h, img_w]
+                }
+            }
+        self.preprocess_op = create_operators(pre_process_list)
+
+    def order_points_clockwise(self, pts):
+        rect = np.zeros((4, 2), dtype="float32")
+        s = pts.sum(axis=1)
+        rect[0] = pts[np.argmin(s)]
+        rect[2] = pts[np.argmax(s)]
+        tmp = np.delete(pts, (np.argmin(s), np.argmax(s)), axis=0)
+        diff = np.diff(np.array(tmp), axis=1)
+        rect[1] = tmp[np.argmin(diff)]
+        rect[3] = tmp[np.argmax(diff)]
+        return rect
+
+    def clip_det_res(self, points, img_height, img_width):
+        for pno in range(points.shape[0]):
+            points[pno, 0] = int(min(max(points[pno, 0], 0), img_width - 1))
+            points[pno, 1] = int(min(max(points[pno, 1], 0), img_height - 1))
+        return points
+
+    def filter_tag_det_res(self, dt_boxes, image_shape):
+        img_height, img_width = image_shape[0:2]
+        dt_boxes_new = []
+        for box in dt_boxes:
+            if isinstance(box, list):
+                box = np.array(box)
+            box = self.order_points_clockwise(box)
+            box = self.clip_det_res(box, img_height, img_width)
+            rect_width = int(np.linalg.norm(box[0] - box[1]))
+            rect_height = int(np.linalg.norm(box[0] - box[3]))
+            if rect_width <= 3 or rect_height <= 3:
+                continue
+            dt_boxes_new.append(box)
+        dt_boxes = np.array(dt_boxes_new)
+        return dt_boxes
+
+    def filter_tag_det_res_only_clip(self, dt_boxes, image_shape):
+        img_height, img_width = image_shape[0:2]
+        dt_boxes_new = []
+        for box in dt_boxes:
+            if isinstance(box, list):
+                box = np.array(box)
+            box = self.clip_det_res(box, img_height, img_width)
+            dt_boxes_new.append(box)
+        dt_boxes = np.array(dt_boxes_new)
+        return dt_boxes
+
+    def __call__(self, img):
+        ori_im = img.copy()
+        data = {'image': img}
+
+        st = time.time()
+        data = transform(data, self.preprocess_op)
+        img, shape_list = data
+        if img is None:
+            return None, 0
+        img = np.expand_dims(img, axis=0)
+        shape_list = np.expand_dims(shape_list, axis=0)
+        img = img.copy()
+        input_dict = {}
+        input_dict[self.input_tensor.name] = img
+        for i in range(100000):
+            try:
+                outputs = self.predictor.run(None, input_dict)
+                break
+            except Exception as e:
+                if i >= 3:
+                    raise e
+                time.sleep(5)
+
+        post_result = self.postprocess_op({"maps": outputs[0]}, shape_list)
+        dt_boxes = post_result[0]['points']
+        dt_boxes = self.filter_tag_det_res(dt_boxes, ori_im.shape)
+
+        return dt_boxes, time.time() - st
+
+
+class RagFlow():
+    def __init__(self, model_dir=None):
+
+        if not model_dir:
+            try:
+                model_dir = os.path.join(
+                        get_deepdoc_directory(),
+                        "models")
+                self.text_detector = RagFlowTextDetector(model_dir)
+                
+
+            except Exception as e:
+                model_dir = snapshot_download(repo_id="InfiniFlow/deepdoc",
+                                              local_dir=os.path.join(get_deepdoc_directory(), "models"),
+                                              local_dir_use_symlinks=False)
+                self.text_detector = RagFlowTextDetector(model_dir)
+
+
+        self.drop_score = 0.5
+        self.crop_image_res_index = 0
+
+    def get_rotate_crop_image(self, img, points):
+        '''
+        img_height, img_width = img.shape[0:2]
+        left = int(np.min(points[:, 0]))
+        right = int(np.max(points[:, 0]))
+        top = int(np.min(points[:, 1]))
+        bottom = int(np.max(points[:, 1]))
+        img_crop = img[top:bottom, left:right, :].copy()
+        points[:, 0] = points[:, 0] - left
+        points[:, 1] = points[:, 1] - top
+        '''
+        assert len(points) == 4, "shape of points must be 4*2"
+        img_crop_width = int(
+            max(
+                np.linalg.norm(points[0] - points[1]),
+                np.linalg.norm(points[2] - points[3])))
+        img_crop_height = int(
+            max(
+                np.linalg.norm(points[0] - points[3]),
+                np.linalg.norm(points[1] - points[2])))
+        pts_std = np.float32([[0, 0], [img_crop_width, 0],
+                              [img_crop_width, img_crop_height],
+                              [0, img_crop_height]])
+        M = cv2.getPerspectiveTransform(points, pts_std)
+        dst_img = cv2.warpPerspective(
+            img,
+            M, (img_crop_width, img_crop_height),
+            borderMode=cv2.BORDER_REPLICATE,
+            flags=cv2.INTER_CUBIC)
+        dst_img_height, dst_img_width = dst_img.shape[0:2]
+        if dst_img_height * 1.0 / dst_img_width >= 1.5:
+            dst_img = np.rot90(dst_img)
+        return dst_img
+
+    def sorted_boxes(self, dt_boxes):
+        """
+        Sort text boxes in order from top to bottom, left to right
+        args:
+            dt_boxes(array):detected text boxes with shape [4, 2]
+        return:
+            sorted boxes(array) with shape [4, 2]
+        """
+        num_boxes = dt_boxes.shape[0]
+        sorted_boxes = sorted(dt_boxes, key=lambda x: (x[0][1], x[0][0]))
+        _boxes = list(sorted_boxes)
+
+        for i in range(num_boxes - 1):
+            for j in range(i, -1, -1):
+                if abs(_boxes[j + 1][0][1] - _boxes[j][0][1]) < 10 and \
+                        (_boxes[j + 1][0][0] < _boxes[j][0][0]):
+                    tmp = _boxes[j]
+                    _boxes[j] = _boxes[j + 1]
+                    _boxes[j + 1] = tmp
+                else:
+                    break
+        return _boxes
+
+    def detect(self, img):
+        time_dict = {'det': 0, 'rec': 0, 'cls': 0, 'all': 0}
+
+        if img is None:
+            return None, None, time_dict
+
+        start = time.time()
+        dt_boxes, elapse = self.text_detector(img)
+        time_dict['det'] = elapse
+
+
+        return zip(self.sorted_boxes(dt_boxes), [
+                   ("", 0) for _ in range(len(dt_boxes))])
+
+    def recognize(self, ori_im, box):
+        img_crop = self.get_rotate_crop_image(ori_im, box)
+
+        rec_res, elapse = self.text_recognizer([img_crop])
+        text, score = rec_res[0]
+        if score < self.drop_score:
+            return ""
+        return text
+
+    def predict(self,img:np.ndarray=None)-> List[List[float]]:
+        """
+        Return np array of bounding boxes - for each box 4 points of 2 coordinates
+        """
+        time_dict = {'det': 0, 'rec': 0, 'cls': 0, 'all': 0}
+
+        dt_boxes, elapse = self.text_detector(img)
+        time_dict['det'] = elapse
+
+
+        dt_boxes = self.sorted_boxes(dt_boxes)
+            
+
+        return dt_boxes
+
+

detectionAndOcrTable1.py

@@ -0,0 +1,425 @@
+from typing import Tuple, List, Sequence, Optional, Union
+from torchvision import transforms
+from torch import nn, Tensor
+from PIL import Image
+from pathlib import Path
+from bs4 import BeautifulSoup as bs
+
+import numpy as np
+import numpy.typing as npt
+from numpy import uint8
+ImageType = npt.NDArray[uint8]
+from transformers import AutoModelForObjectDetection
+import torch
+import matplotlib.pyplot as plt
+import matplotlib.patches as patches
+from matplotlib.patches import Patch
+
+from unitable import UnitablePredictor
+from doctrfiles import DoctrWordDetector,DoctrTextRecognizer
+from utils import crop_an_Image,cropImageExtraMargin
+from utils import denoisingAndSharpening
+
+#based on this notebook:https://github.com/NielsRogge/Transformers-Tutorials/blob/master/Table%20Transformer/Inference_with_Table_Transformer_(TATR)_for_parsing_tables.ipynb
+class MaxResize(object):
+    def __init__(self, max_size=800):
+        self.max_size = max_size
+
+    def __call__(self, image):
+        width, height = image.size
+        current_max_size = max(width, height)
+        scale = self.max_size / current_max_size
+        resized_image = image.resize((int(round(scale*width)), int(round(scale*height))))
+
+        return resized_image
+    
+
+html_table_template = (
+
+    lambda table: f"""<html>
+        <head> <meta charset="UTF-8">
+        <style>
+        table, th, td {{
+            border: 1px solid black;
+            font-size: 10px;
+        }}
+        </style> </head>
+        <body>
+        <table frame="hsides" rules="groups" width="100%%">
+            {table}
+        </table> </body> </html>"""
+)
+
+class DetectionAndOcrTable1():
+    def __init__(self,englishFlag=True):
+        self.unitablePredictor = UnitablePredictor()
+        self.wordDetector = DoctrWordDetector(architecture="db_resnet50", 
+                                              path_weights="doctrfiles/models/db_resnet50-79bd7d70.pt",
+                                              path_config_json ="doctrfiles/models/db_resnet50_config.json")
+
+
+        if englishFlag:
+            self.textRecognizer = DoctrTextRecognizer(architecture="master", path_weights="./doctrfiles/models/master-fde31e4a.pt", 
+                                            path_config_json="./doctrfiles/models/master.json")
+        else:
+            self.textRecognizer = DoctrTextRecognizer(architecture="parseq", path_weights="./doctrfiles/models/doctr-multilingual-parseq.bin", 
+                                           path_config_json="./doctrfiles/models/multilingual-parseq-config.json")
+        
+
+    @staticmethod
+    def build_table_from_html_and_cell(
+        structure: List[str], content: List[str] = None
+    ) -> List[str]:
+        """Build table from html and cell token list"""
+        assert structure is not None
+        html_code = list()
+
+        # deal with empty table
+        if content is None:
+            content = ["placeholder"] * len(structure)
+
+        for tag in structure:
+            if tag in ("<td>[]</td>", ">[]</td>"):
+                if len(content) == 0:
+                    continue
+                cell = content.pop(0)
+                html_code.append(tag.replace("[]", cell))
+            else:
+                html_code.append(tag)
+
+        return html_code
+    
+    @staticmethod
+    def save_detection(detected_lines_images:List[ImageType], prefix = './res/test1/res_'):
+        i = 0
+        for img in detected_lines_images:
+            pilimg = Image.fromarray(img)
+            pilimg.save(prefix+str(i)+'.png')
+            i=i+1
+        
+    @staticmethod
+    # for output bounding box post-processing
+    def box_cxcywh_to_xyxy(x):
+        x_c, y_c, w, h = x.unbind(-1)
+        b = [(x_c - 0.5 * w), (y_c - 0.5 * h), (x_c + 0.5 * w), (y_c + 0.5 * h)]
+        return torch.stack(b, dim=1)
+
+    @staticmethod
+    def rescale_bboxes(out_bbox, size):
+        img_w, img_h = size
+        b = DetectionAndOcrTable1.box_cxcywh_to_xyxy(out_bbox)
+        b = b * torch.tensor([img_w, img_h, img_w, img_h], dtype=torch.float32)
+        return b
+
+    @staticmethod
+    def outputs_to_objects(outputs, img_size, id2label):
+        m = outputs.logits.softmax(-1).max(-1)
+        pred_labels = list(m.indices.detach().cpu().numpy())[0]
+        pred_scores = list(m.values.detach().cpu().numpy())[0]
+        pred_bboxes = outputs['pred_boxes'].detach().cpu()[0]
+        pred_bboxes = [elem.tolist() for elem in DetectionAndOcrTable1.rescale_bboxes(pred_bboxes, img_size)]
+
+        objects = []
+        for label, score, bbox in zip(pred_labels, pred_scores, pred_bboxes):
+            class_label = id2label[int(label)]
+            if not class_label == 'no object':
+                objects.append({'label': class_label, 'score': float(score),
+                                'bbox': [float(elem) for elem in bbox]})
+
+        return objects
+    
+    @staticmethod
+    def fig2img(fig):
+        """Convert a Matplotlib figure to a PIL Image and return it"""
+        import io
+        buf = io.BytesIO()
+        fig.savefig(buf)
+        buf.seek(0)
+        img = Image.open(buf)
+        return img
+        #For that, the TATR authors employ some padding to make sure the borders of the table are included.
+        
+    @staticmethod
+    def objects_to_crops(img, tokens, objects, class_thresholds, padding=10):
+        """
+        Process the bounding boxes produced by the table detection model into
+        cropped table images and cropped tokens.
+        """
+
+        table_crops = []
+        for obj in objects:
+            # abit unecessary here cause i crop them anywyas 
+            if obj['score'] < class_thresholds[obj['label']]:
+                continue
+
+            cropped_table = {}
+
+            bbox = obj['bbox']
+            bbox = [bbox[0]-padding, bbox[1]-padding, bbox[2]+padding, bbox[3]+padding]
+
+            cropped_img = img.crop(bbox)
+            
+            # Add padding to the cropped image
+            padded_width = cropped_img.width + 40
+            padded_height = cropped_img.height +40
+            
+            new_img_np = np.full((padded_height, padded_width, 3), fill_value=255, dtype=np.uint8)
+            y_offset = (padded_height - cropped_img.height) // 2
+            x_offset = (padded_width - cropped_img.width) // 2
+            new_img_np[y_offset:y_offset + cropped_img.height, x_offset:x_offset+cropped_img.width] = np.array(cropped_img)
+
+            padded_img = Image.fromarray(new_img_np,'RGB')
+
+
+            table_tokens = [token for token in tokens if iob(token['bbox'], bbox) >= 0.5]
+            for token in table_tokens:
+                token['bbox'] = [token['bbox'][0]-bbox[0] + padding,
+                                 token['bbox'][1]-bbox[1] + padding,
+                                 token['bbox'][2]-bbox[0] + padding,
+                                 token['bbox'][3]-bbox[1] + padding]
+
+            # If table is predicted to be rotated, rotate cropped image and tokens/words:
+            if obj['label'] == 'table rotated':
+                padded_img = padded_img.rotate(270, expand=True)
+                for token in table_tokens:
+                    bbox = token['bbox']
+                    bbox = [padded_img.size[0]-bbox[3]-1,
+                            bbox[0],
+                            padded_img.size[0]-bbox[1]-1,
+                            bbox[2]]
+                    token['bbox'] = bbox
+
+            cropped_table['image'] = padded_img
+            cropped_table['tokens'] = table_tokens
+
+            table_crops.append(cropped_table)
+
+        return table_crops
+    
+    @staticmethod
+    def visualize_detected_tables(img, det_tables, out_path=None):
+        plt.imshow(img, interpolation="lanczos")
+        fig = plt.gcf()
+        fig.set_size_inches(20, 20)
+        ax = plt.gca()
+
+        for det_table in det_tables:
+            bbox = det_table['bbox']
+
+            if det_table['label'] == 'table':
+                facecolor = (1, 0, 0.45)
+                edgecolor = (1, 0, 0.45)
+                alpha = 0.3
+                linewidth = 2
+                hatch='//////'
+            elif det_table['label'] == 'table rotated':
+                facecolor = (0.95, 0.6, 0.1)
+                edgecolor = (0.95, 0.6, 0.1)
+                alpha = 0.3
+                linewidth = 2
+                hatch='//////'
+            else:
+                continue
+
+            rect = patches.Rectangle(bbox[:2], bbox[2]-bbox[0], bbox[3]-bbox[1], linewidth=linewidth,
+                                        edgecolor='none',facecolor=facecolor, alpha=0.1)
+            ax.add_patch(rect)
+            rect = patches.Rectangle(bbox[:2], bbox[2]-bbox[0], bbox[3]-bbox[1], linewidth=linewidth,
+                                        edgecolor=edgecolor,facecolor='none',linestyle='-', alpha=alpha)
+            ax.add_patch(rect)
+            rect = patches.Rectangle(bbox[:2], bbox[2]-bbox[0], bbox[3]-bbox[1], linewidth=0,
+                                        edgecolor=edgecolor,facecolor='none',linestyle='-', hatch=hatch, alpha=0.2)
+            ax.add_patch(rect)
+
+        plt.xticks([], [])
+        plt.yticks([], [])
+
+        legend_elements = [Patch(facecolor=(1, 0, 0.45), edgecolor=(1, 0, 0.45),
+                                    label='Table', hatch='//////', alpha=0.3),
+                            Patch(facecolor=(0.95, 0.6, 0.1), edgecolor=(0.95, 0.6, 0.1),
+                                    label='Table (rotated)', hatch='//////', alpha=0.3)]
+        plt.legend(handles=legend_elements, bbox_to_anchor=(0.5, -0.02), loc='upper center', borderaxespad=0,
+                        fontsize=10, ncol=2)
+        plt.gcf().set_size_inches(10, 10)
+        plt.axis('off')
+
+        if out_path is not None:
+            plt.savefig(out_path, bbox_inches='tight', dpi=150)
+
+        return fig
+    
+    
+    def predict(self,image:Image.Image,debugfolder_filename_page_name,denoise=False):
+        
+        
+        """
+        0. Locate the table using Table detection 
+        1. Unitable 
+        """
+        print("Running table transformer + Unitable Hybrid Model")
+
+        # Step 0 : Locate the table using Table detection TODO
+
+        #First we load a Table Transformer pre-trained for table detection. We use the "no_timm" version here to load the checkpoint with a Transformers-native backbone.
+        model = AutoModelForObjectDetection.from_pretrained("microsoft/table-transformer-detection", revision="no_timm")
+        device = "cuda" if torch.cuda.is_available() else "cpu"
+        model.to(device)
+
+        #Preparing the image for the model 
+        detection_transform = transforms.Compose([
+            MaxResize(800),
+            transforms.ToTensor(),
+            transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
+        ])
+        pixel_values = detection_transform(image).unsqueeze(0)
+        pixel_values = pixel_values.to(device)
+        
+        # Next, we forward the pixel values through the model. 
+        # The model outputs logits of shape (batch_size, num_queries, num_labels + 1). The +1 is for the "no object" class.
+        with torch.no_grad():
+            outputs = model(pixel_values)
+
+        # update id2label to include "no object"
+        id2label = model.config.id2label
+        id2label[len(model.config.id2label)] = "no object"
+
+        #[{'label': 'table', 'score': 0.9999570846557617, 'bbox': [110.24547576904297, 73.31171417236328, 1024.609130859375, 308.7159423828125]}]
+        objects = DetectionAndOcrTable1.outputs_to_objects(outputs, image.size, id2label)
+        
+        #Only do these for objects with score greater than 0.8 
+        objects = [obj for obj in objects if obj['score'] > 0.95]
+
+        print("detected object from the table transformers are")
+        print(objects)
+        if objects: 
+            
+            #Next, we crop the table out of the image. For that, the TATR authors employ some padding to make sure the borders of the table are included.
+
+
+            tokens = []
+            detection_class_thresholds = {
+                "table": 0.95, #this is a bit double cause we do up there another filtering but didn't want to modify too much from original code 
+                "table rotated": 0.95,
+                "no object": 10
+            }
+            crop_padding = 10
+
+            
+            tables_crops = DetectionAndOcrTable1.objects_to_crops(image, tokens, objects, detection_class_thresholds, padding=crop_padding)
+
+            cropped_tables =[]
+            for i in range (len(tables_crops)):
+                cropped_table = tables_crops[i]['image'].convert("RGB")
+                cropped_table.save(debugfolder_filename_page_name+"cropped_table_"+str(i)+".png")
+                cropped_tables.append(cropped_table)
+
+            # Step 1: Unitable 
+            #This take PIL Images as input 
+            if denoise:
+                cropped_tables =denoisingAndSharpening(cropped_tables)
+            pred_htmls, pred_bboxs = self.unitablePredictor.predict(cropped_tables,debugfolder_filename_page_name)
+            
+            table_codes = []
+            for k in range(len(cropped_tables)):
+                pred_html =pred_htmls[k]
+                pred_bbox = pred_bboxs[k]
+
+                # Some tabless have a lot of words in their header 
+                # So for the headers, give doctr word ddetector doesn't work when the images aren't square 
+                table_header_cells = 0
+                header_exists = False
+                for cell in pred_html:
+                    if cell=='>[]</td>' or cell == '<td>[]</td>':
+                        table_header_cells += 1
+                    if cell =='</thead>':
+                        header_exists = True
+                        break
+                if not header_exists:
+                    table_header_cells = 0
+                pred_cell = []
+                cell_imgs_to_viz = []
+                cell_img_num=0
+                
+                # Find what one line should be if there is a cell with a single line 
+                one_line_height = 100000
+                for i in range(table_header_cells):
+                    box = pred_bbox[i]
+                    xmin, ymin, xmax, ymax = box
+                    current_box_height = abs(ymax-ymin) 
+                    if current_box_height<one_line_height:
+                        one_line_height = current_box_height
+
+                for box in pred_bbox:
+                    xmin, ymin, xmax, ymax = box
+                    fourbytwo = np.array([
+                            [xmin, ymin],
+                            [xmax, ymin],
+                            [xmax, ymax],
+                            [xmin, ymax]
+                        ], dtype=np.float32)
+                    current_box_height = abs(ymax-ymin) 
+                    
+                    # Those are for header cells with more than one line
+                    if table_header_cells > 0 and current_box_height>one_line_height+5:
+
+                        cell_img= cropImageExtraMargin([fourbytwo],cropped_tables[k],margin=1.4)[0]
+                        table_header_cells -= 1
+
+                         #List of 4 x 2 
+                        detection_results = self.wordDetector.predict(cell_img,sort_vertical=True)
+
+                        input_to_recog = []
+                        if detection_results == []:
+                            input_to_recog.append(cell_img)
+                        else: 
+
+                            for wordbox in detection_results:
+
+                                cropped_image= crop_an_Image(wordbox.box,cell_img)
+                                if cropped_image.shape[0] >0 and cropped_image.shape[1]>0:
+                                    input_to_recog.append(cropped_image)
+                                else: 
+                                    print("Empty image")
+                    else:
+                        cell_img = crop_an_Image(fourbytwo,cropped_tables[k])
+                        if table_header_cells>0:
+                            table_header_cells -= 1
+                        if cell_img.shape[0] >0 and cell_img.shape[1]>0:
+                            input_to_recog =[cell_img]
+
+                    cell_imgs_to_viz.append(cell_img)
+                    
+
+                    if input_to_recog != []:
+                        words = self.textRecognizer.predict_for_tables(input_to_recog)
+                        cell_output = " ".join(words)
+                        pred_cell.append(cell_output)
+                    else:
+                        #Don't lose empty cell 
+                        pred_cell.append("")
+
+
+                print(pred_cell)
+                #Step3 : 
+                pred_code = self.build_table_from_html_and_cell(pred_html, pred_cell)
+                pred_code = "".join(pred_code)
+                pred_code = html_table_template(pred_code)
+
+
+                soup = bs(pred_code)
+                #formatted and indented) string representation of the HTML document
+                table_code = soup.prettify()
+                print(table_code)
+
+                # Append extracted table to table_codes 
+                table_codes.append(table_code)
+
+            return table_codes
+
+
+
+        
+
+
+
+        

detectionAndOcrTable2.py

@@ -0,0 +1,306 @@
+from typing import Tuple, List, Sequence, Optional, Union
+from torchvision import transforms
+from torch import nn, Tensor
+from PIL import Image
+from pathlib import Path
+from bs4 import BeautifulSoup as bs
+
+import numpy as np
+import numpy.typing as npt
+from numpy import uint8
+ImageType = npt.NDArray[uint8]
+from transformers import AutoModelForObjectDetection
+import torch
+import matplotlib.pyplot as plt
+import matplotlib.patches as patches
+from matplotlib.patches import Patch
+
+from unitable import UnitableFullPredictor
+
+#based on this notebook:https://github.com/NielsRogge/Transformers-Tutorials/blob/master/Table%20Transformer/Inference_with_Table_Transformer_(TATR)_for_parsing_tables.ipynb
+class MaxResize(object):
+    def __init__(self, max_size=800):
+        self.max_size = max_size
+
+    def __call__(self, image):
+        width, height = image.size
+        current_max_size = max(width, height)
+        scale = self.max_size / current_max_size
+        resized_image = image.resize((int(round(scale*width)), int(round(scale*height))))
+
+        return resized_image
+    
+def iob(boxA, boxB):
+    """
+    Calculate the Intersection over Bounding Box (IoB) of two bounding boxes.
+    
+    Parameters:
+    - boxA: list or tuple with [xmin, ymin, xmax, ymax] of the first box
+    - boxB: list or tuple with [xmin, ymin, xmax, ymax] of the second box
+    
+    Returns:
+    - iob: float, the IoB ratio
+    """
+    # Determine the coordinates of the intersection rectangle
+    xA = max(boxA[0], boxB[0])
+    yA = max(boxA[1], boxB[1])
+    xB = min(boxA[2], boxB[2])
+    yB = min(boxA[3], boxB[3])
+    
+    # Compute the area of intersection rectangle
+    interWidth = max(0, xB - xA)
+    interHeight = max(0, yB - yA)
+    interArea = interWidth * interHeight
+    
+    # Compute the area of boxB (the second box)
+    boxBArea = (boxB[2] - boxB[0]) * (boxB[3] - boxB[1])
+    
+    # Compute the Intersection over Bounding Box (IoB) ratio
+    iob = interArea / float(boxBArea)
+    
+    return iob
+
+class DetectionAndOcrTable2():
+    #This components can take in entire pdf page as input , scan for tables and return the table in html format
+    #Uses the full unitable model - different to DetectionAndOcrTable1
+    def __init__(self):
+        self.unitableFullPredictor = UnitableFullPredictor()
+
+    
+    @staticmethod
+    def save_detection(detected_lines_images:List[ImageType], prefix = './res/test1/res_'):
+        i = 0
+        for img in detected_lines_images:
+            pilimg = Image.fromarray(img)
+            pilimg.save(prefix+str(i)+'.png')
+            i=i+1
+    
+    @staticmethod
+    # for output bounding box post-processing
+    def box_cxcywh_to_xyxy(x):
+        x_c, y_c, w, h = x.unbind(-1)
+        b = [(x_c - 0.5 * w), (y_c - 0.5 * h), (x_c + 0.5 * w), (y_c + 0.5 * h)]
+        return torch.stack(b, dim=1)
+
+    @staticmethod
+    def rescale_bboxes(out_bbox, size):
+        img_w, img_h = size
+        b = DetectionAndOcrTable2.box_cxcywh_to_xyxy(out_bbox)
+        b = b * torch.tensor([img_w, img_h, img_w, img_h], dtype=torch.float32)
+        return b
+
+    @staticmethod
+    def outputs_to_objects(outputs, img_size, id2label):
+        m = outputs.logits.softmax(-1).max(-1)
+        pred_labels = list(m.indices.detach().cpu().numpy())[0]
+        pred_scores = list(m.values.detach().cpu().numpy())[0]
+        pred_bboxes = outputs['pred_boxes'].detach().cpu()[0]
+        pred_bboxes = [elem.tolist() for elem in DetectionAndOcrTable2.rescale_bboxes(pred_bboxes, img_size)]
+
+        objects = []
+        for label, score, bbox in zip(pred_labels, pred_scores, pred_bboxes):
+            class_label = id2label[int(label)]
+            if not class_label == 'no object':
+                objects.append({'label': class_label, 'score': float(score),
+                                'bbox': [float(elem) for elem in bbox]})
+
+        return objects
+    
+
+    @staticmethod
+    def visualize_detected_tables(img, det_tables, out_path=None):
+        plt.imshow(img, interpolation="lanczos")
+        fig = plt.gcf()
+        fig.set_size_inches(20, 20)
+        ax = plt.gca()
+
+        for det_table in det_tables:
+            bbox = det_table['bbox']
+
+            if det_table['label'] == 'table':
+                facecolor = (1, 0, 0.45)
+                edgecolor = (1, 0, 0.45)
+                alpha = 0.3
+                linewidth = 2
+                hatch='//////'
+            elif det_table['label'] == 'table rotated':
+                facecolor = (0.95, 0.6, 0.1)
+                edgecolor = (0.95, 0.6, 0.1)
+                alpha = 0.3
+                linewidth = 2
+                hatch='//////'
+            else:
+                continue
+
+            rect = patches.Rectangle(bbox[:2], bbox[2]-bbox[0], bbox[3]-bbox[1], linewidth=linewidth,
+                                        edgecolor='none',facecolor=facecolor, alpha=0.1)
+            ax.add_patch(rect)
+            rect = patches.Rectangle(bbox[:2], bbox[2]-bbox[0], bbox[3]-bbox[1], linewidth=linewidth,
+                                        edgecolor=edgecolor,facecolor='none',linestyle='-', alpha=alpha)
+            ax.add_patch(rect)
+            rect = patches.Rectangle(bbox[:2], bbox[2]-bbox[0], bbox[3]-bbox[1], linewidth=0,
+                                        edgecolor=edgecolor,facecolor='none',linestyle='-', hatch=hatch, alpha=0.2)
+            ax.add_patch(rect)
+
+        plt.xticks([], [])
+        plt.yticks([], [])
+
+        legend_elements = [Patch(facecolor=(1, 0, 0.45), edgecolor=(1, 0, 0.45),
+                                    label='Table', hatch='//////', alpha=0.3),
+                            Patch(facecolor=(0.95, 0.6, 0.1), edgecolor=(0.95, 0.6, 0.1),
+                                    label='Table (rotated)', hatch='//////', alpha=0.3)]
+        plt.legend(handles=legend_elements, bbox_to_anchor=(0.5, -0.02), loc='upper center', borderaxespad=0,
+                        fontsize=10, ncol=2)
+        plt.gcf().set_size_inches(10, 10)
+        plt.axis('off')
+
+        if out_path is not None:
+            plt.savefig(out_path, bbox_inches='tight', dpi=150)
+
+        return fig
+    
+    #For that, the TATR authors employ some padding to make sure the borders of the table are included.
+    @staticmethod
+    def objects_to_crops(img, tokens, objects, class_thresholds, padding=10):
+        """
+        Process the bounding boxes produced by the table detection model into
+        cropped table images and cropped tokens.
+        """
+
+        table_crops = []
+        for obj in objects:
+            # abit unecessary here cause i crop them anywyas 
+            if obj['score'] < class_thresholds[obj['label']]:
+                print('skipping object with score', obj['score'])
+                continue
+
+            cropped_table = {}
+
+            bbox = obj['bbox']
+            bbox = [bbox[0]-padding, bbox[1]-padding, bbox[2]+padding, bbox[3]+padding]
+
+            cropped_img = img.crop(bbox)
+            
+            # Add padding to the cropped image
+            padded_width = cropped_img.width + 40
+            padded_height = cropped_img.height +40
+            
+            new_img_np = np.full((padded_height, padded_width, 3), fill_value=255, dtype=np.uint8)
+            y_offset = (padded_height - cropped_img.height) // 2
+            x_offset = (padded_width - cropped_img.width) // 2
+            new_img_np[y_offset:y_offset + cropped_img.height, x_offset:x_offset+cropped_img.width] = np.array(cropped_img)
+
+            padded_img = Image.fromarray(new_img_np,'RGB')
+
+
+            table_tokens = [token for token in tokens if iob(token['bbox'], bbox) >= 0.5]
+            for token in table_tokens:
+                token['bbox'] = [token['bbox'][0]-bbox[0] + padding,
+                                 token['bbox'][1]-bbox[1] + padding,
+                                 token['bbox'][2]-bbox[0] + padding,
+                                 token['bbox'][3]-bbox[1] + padding]
+
+            # If table is predicted to be rotated, rotate cropped image and tokens/words:
+            if obj['label'] == 'table rotated':
+                padded_img = padded_img.rotate(270, expand=True)
+                for token in table_tokens:
+                    bbox = token['bbox']
+                    bbox = [padded_img.size[0]-bbox[3]-1,
+                            bbox[0],
+                            padded_img.size[0]-bbox[1]-1,
+                            bbox[2]]
+                    token['bbox'] = bbox
+
+            cropped_table['image'] = padded_img
+            cropped_table['tokens'] = table_tokens
+
+            table_crops.append(cropped_table)
+
+        return table_crops
+    
+    def predict(self,image:Image.Image,debugfolder_filename_page_name):
+        
+        
+        """
+        0. Locate the table using Table detection 
+        1. Unitable 
+        """
+
+        # Step 0 : Locate the table using Table detection TODO
+
+        #First we load a Table Transformer pre-trained for table detection. We use the "no_timm" version here to load the checkpoint with a Transformers-native backbone.
+        model = AutoModelForObjectDetection.from_pretrained("microsoft/table-transformer-detection", revision="no_timm")
+        device = "cuda" if torch.cuda.is_available() else "cpu"
+        model.to(device)
+
+        #Preparing the image for the model 
+        detection_transform = transforms.Compose([
+            MaxResize(800),
+            transforms.ToTensor(),
+            transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
+        ])
+        pixel_values = detection_transform(image).unsqueeze(0)
+        pixel_values = pixel_values.to(device)
+        
+        # Next, we forward the pixel values through the model. 
+        # The model outputs logits of shape (batch_size, num_queries, num_labels + 1). The +1 is for the "no object" class.
+        with torch.no_grad():
+            outputs = model(pixel_values)
+
+        # update id2label to include "no object"
+        id2label = model.config.id2label
+        id2label[len(model.config.id2label)] = "no object"
+
+        #[{'label': 'table', 'score': 0.9999570846557617, 'bbox': [110.24547576904297, 73.31171417236328, 1024.609130859375, 308.7159423828125]}]
+        objects = DetectionAndOcrTable2.outputs_to_objects(outputs, image.size, id2label)
+        
+        #Only do these for objects with score greater than 0.8 
+        objects = [obj for obj in objects if obj['score'] > 0.95]
+
+        print(objects)
+        if objects: 
+            fig = DetectionAndOcrTable2.visualize_detected_tables(image, objects,out_path = "./res/table_debug/table_former_detection.jpg")
+
+            #Next, we crop the table out of the image. For that, the TATR authors employ some padding to make sure the borders of the table are included.
+
+
+            tokens = []
+            detection_class_thresholds = {
+                "table": 0.95,
+                "table rotated": 0.95,
+                "no object": 10
+            }
+            crop_padding = 10
+
+            
+            tables_crops = DetectionAndOcrTable2.objects_to_crops(image, tokens, objects, detection_class_thresholds, padding=crop_padding)
+            
+            #[{'image': <PIL.Image.Image image mode=RGB size=1392x903 at 0x7F71B02BCB50>, 'tokens': []}]
+            #print(tables_crops)
+
+            #TODO: Handle the case where there are multiple tables 
+            cropped_tables =[]
+            for i in range (len(tables_crops)):
+                cropped_table = tables_crops[i]['image'].convert("RGB")
+                cropped_table.save(debugfolder_filename_page_name +"cropped_table_"+str(i)+".png")
+                cropped_tables.append(cropped_table)
+
+            print("number of cropped tables found: "+str(len(cropped_tables)))
+
+            
+            # Step 1: Unitable 
+            #This take PIL Images as input 
+            table_codes = self.unitableFullPredictor.predict(cropped_tables,debugfolder_filename_page_name)
+
+        else: 
+            return 
+
+        
+
+
+
+        
+
+
+
+        

detectionAndOcrTable3.py

@@ -0,0 +1,267 @@
+from typing import Tuple, List, Sequence, Optional, Union
+from torchvision import transforms
+from torch import nn, Tensor
+from PIL import Image
+from pathlib import Path
+from bs4 import BeautifulSoup as bs
+
+import numpy as np
+import numpy.typing as npt
+from numpy import uint8
+ImageType = npt.NDArray[uint8]
+from transformers import AutoModelForObjectDetection
+import torch
+import matplotlib.pyplot as plt
+import matplotlib.patches as patches
+from matplotlib.patches import Patch
+from utils import draw_only_box
+
+from unitable import UnitablePredictor
+from ultralyticsplus import YOLO, render_result
+from doctrfiles import DoctrWordDetector,DoctrTextRecognizer
+from utils import crop_an_Image,cropImageExtraMargin
+from utils import denoisingAndSharpening
+"""
+USES YOLO FOR DETECITON INSTEAD OF TABLE TRANSFORMER 
+Table TransFORMER
+"""
+
+
+html_table_template = (
+
+    lambda table: f"""<html>
+        <head> <meta charset="UTF-8">
+        <style>
+        table, th, td {{
+            border: 1px solid black;
+            font-size: 10px;
+        }}
+        </style> </head>
+        <body>
+        <table frame="hsides" rules="groups" width="100%%">
+            {table}
+        </table> </body> </html>"""
+)
+
+class DetectionAndOcrTable3():
+    #This components can take in entire pdf page as input , scan for tables and return the table in html format
+    #Uses the full unitable model - different to DetectionAndOcrTable1
+    def __init__(self,englishFlag = True):
+        self.unitablePredictor = UnitablePredictor()
+        self.detector = YOLO('foduucom/table-detection-and-extraction')
+        # set model parameters
+        self.detector.overrides['conf'] = 0.25 # NMS confidence threshold
+        self.detector.overrides['iou'] = 0.45  # NMS IoU threshold
+        self.detector.overrides['agnostic_nms'] = False  # NMS class-agnostic
+        self.detector.overrides['max_det'] = 1000  # maximum number of detections per image
+
+        self.wordDetector = DoctrWordDetector(architecture="db_resnet50", 
+                                              path_weights="doctrfiles/models/db_resnet50-79bd7d70.pt",
+                                              path_config_json ="doctrfiles/models/db_resnet50_config.json")
+
+
+        if englishFlag:
+            self.textRecognizer = DoctrTextRecognizer(architecture="master", path_weights="./doctrfiles/models/master-fde31e4a.pt", 
+                                            path_config_json="./doctrfiles/models/master.json")
+        else:
+            self.textRecognizer = DoctrTextRecognizer(architecture="parseq", path_weights="./doctrfiles/models/doctr-multilingual-parseq.bin", 
+                                           path_config_json="./doctrfiles/models/multilingual-parseq-config.json")
+        
+
+    
+    @staticmethod
+    def save_detection(detected_lines_images:List[ImageType], prefix = './res/test1/res_'):
+        i = 0
+        for img in detected_lines_images:
+            pilimg = Image.fromarray(img)
+            pilimg.save(prefix+str(i)+'.png')
+            i=i+1
+
+    @staticmethod
+    def build_table_from_html_and_cell(
+        structure: List[str], content: List[str] = None
+    ) -> List[str]:
+        """Build table from html and cell token list"""
+        assert structure is not None
+        html_code = list()
+
+        # deal with empty table
+        if content is None:
+            content = ["placeholder"] * len(structure)
+
+        for tag in structure:
+            if tag in ("<td>[]</td>", ">[]</td>"):
+                if len(content) == 0:
+                    continue
+                cell = content.pop(0)
+                html_code.append(tag.replace("[]", cell))
+            else:
+                html_code.append(tag)
+
+        return html_code
+    """
+    Valid 'Boxes' object attributes and properties are:
+
+            Attributes:
+                boxes (torch.Tensor) or (numpy.ndarray): A tensor or numpy array containing the detection boxes,
+                    with shape (num_boxes, 6).
+                orig_shape (torch.Tensor) or (numpy.ndarray): Original image size, in the format (height, width).
+
+            Properties:
+                xyxy (torch.Tensor) or (numpy.ndarray): The boxes in xyxy format.
+                conf (torch.Tensor) or (numpy.ndarray): The confidence values of the boxes.
+                cls (torch.Tensor) or (numpy.ndarray): The class values of the boxes.
+                xywh (torch.Tensor) or (numpy.ndarray): The boxes in xywh format.
+                xyxyn (torch.Tensor) or (numpy.ndarray): The boxes in xyxy format normalized by original image size.
+                xywhn (torch.Tensor) or (numpy.ndarray): The boxes in xywh format normalized by original image size.
+    """
+    # Image is page image 
+    def predict(self,image:Image.Image,debugfolder_filename_page_name = None,denoise =False):
+        
+        results = self.detector.predict(image)
+
+        #Array of bboxes
+        bbxs = results[0].boxes.xyxy.int().tolist()
+        #Array of confidences 
+        conf = results[0].boxes.conf.float().tolist()
+        print(bbxs)
+        print(conf)
+    
+        #images_to_recognizer = cropImage(bxs, img)
+        img_to_save = draw_only_box(image, bbxs)
+        img_to_save.save(debugfolder_filename_page_name+"detectionBoxRes.png", quality=95)
+
+        # we need something to draw the detection 
+
+
+        cropped_tables =[]
+        for i in range (len(bbxs)):
+            # TODO: find the right confidence and padding values
+            if conf[i]< 0.65:
+                continue
+        
+            padded = [bbxs[i][0]-10,bbxs[i][1]-10,bbxs[i][2]+10,bbxs[i][3]+10]
+
+            cropped_table = image.convert("RGB").crop(padded)
+            cropped_table.save(debugfolder_filename_page_name +"yolo_cropped_table_"+str(i)+".png")
+            cropped_tables.append(cropped_table)
+
+        print("number of cropped tables found: "+str(len(cropped_tables)))
+        
+        # Step 1: Unitable 
+        #This take PIL Images as input 
+        if cropped_tables != []:
+           if denoise:
+                cropped_tables =denoisingAndSharpening(cropped_tables)
+           pred_htmls, pred_bboxs = self.unitablePredictor.predict(cropped_tables,debugfolder_filename_page_name)
+           table_codes = []
+
+           for k in range(len(cropped_tables)):
+                pred_html =pred_htmls[k]
+                pred_bbox = pred_bboxs[k]
+
+                # Some tabless have a lot of words in their header 
+                # So for the headers, give doctr word ddetector doesn't work when the images aren't square 
+                table_header_cells = 0
+                header_exists = False
+                for cell in pred_html:
+                    if cell=='>[]</td>' or cell == '<td>[]</td>':
+                        table_header_cells += 1
+                    if cell =='</thead>':
+                        header_exists = True
+                        break
+                if not header_exists:
+                    table_header_cells = 0
+                pred_cell = []
+                cell_imgs_to_viz = []
+                cell_img_num=0
+
+                # Find what one line should be if there is a cell with a single line 
+                one_line_height = 100000
+                for i in range(table_header_cells):
+                    box = pred_bbox[i]
+                    xmin, ymin, xmax, ymax = box
+                    current_box_height = abs(ymax-ymin) 
+                    if current_box_height<one_line_height:
+                        one_line_height = current_box_height
+
+                for box in pred_bbox:
+                    xmin, ymin, xmax, ymax = box
+                    fourbytwo = np.array([
+                            [xmin, ymin],
+                            [xmax, ymin],
+                            [xmax, ymax],
+                            [xmin, ymax]
+                        ], dtype=np.float32)
+                    if ymax-ymin == 0:
+                        continue
+                    current_box_height = abs(ymax-ymin) 
+
+                    # Those are for header cells with more than one line
+                    if table_header_cells > 0 and current_box_height>one_line_height+5:
+
+                        cell_img= cropImageExtraMargin([fourbytwo],cropped_tables[k],margin=1.4)[0]
+                        table_header_cells -= 1
+
+                         #List of 4 x 2 
+                        detection_results = self.wordDetector.predict(cell_img,sort_vertical=True)
+
+                        input_to_recog = []
+                        if detection_results == []:
+                            input_to_recog.append(cell_img)
+                        else: 
+
+                            for wordbox in detection_results:
+
+                                cropped_image= crop_an_Image(wordbox.box,cell_img)
+                                if cropped_image.shape[0] >0 and cropped_image.shape[1]>0:
+                                    input_to_recog.append(cropped_image)
+                                else: 
+                                    print("Empty image")
+                    else:
+                        cell_img = crop_an_Image(fourbytwo,cropped_tables[k])
+                        if table_header_cells>0:
+                            table_header_cells -= 1
+                        if cell_img.shape[0] >0 and cell_img.shape[1]>0:
+                            input_to_recog =[cell_img]
+
+                    cell_imgs_to_viz.append(cell_img)
+
+                    if input_to_recog != []:
+                        words = self.textRecognizer.predict_for_tables(input_to_recog)
+                        cell_output = " ".join(words)
+                        pred_cell.append(cell_output)
+                    else:
+                        #Don't lose empty cell 
+                        pred_cell.append("")
+
+
+                #self.save_detection(cell_imgs_to_viz,prefix = './res/test4/cell_imgs_')
+                print(pred_cell)
+                #Step3 : 
+                pred_code = self.build_table_from_html_and_cell(pred_html, pred_cell)
+                pred_code = "".join(pred_code)
+                pred_code = html_table_template(pred_code)
+
+
+                soup = bs(pred_code)
+                #formatted and indented) string representation of the HTML document
+                table_code = soup.prettify()
+                print(table_code)
+                table_codes.append(table_code)
+           
+           return table_codes
+        return []
+
+
+
+
+        
+
+
+
+        
+
+
+
+        

detectionAndOcrTable4.py

@@ -0,0 +1,112 @@
+from typing import Tuple, List, Sequence, Optional, Union
+from torchvision import transforms
+from torch import nn, Tensor
+from PIL import Image
+from pathlib import Path
+from bs4 import BeautifulSoup as bs
+
+import numpy as np
+import numpy.typing as npt
+from numpy import uint8
+ImageType = npt.NDArray[uint8]
+from transformers import AutoModelForObjectDetection
+import torch
+import matplotlib.pyplot as plt
+import matplotlib.patches as patches
+from matplotlib.patches import Patch
+from utils import draw_only_box
+
+from unitable import UnitableFullPredictor
+from ultralyticsplus import YOLO, render_result
+"""
+USES YOLO FOR DETECITON INSTEAD OF TABLE TRANSFORMER 
+Table TransFORMER
+"""
+
+class DetectionAndOcrTable4():
+    #This components can take in entire pdf page as input , scan for tables and return the table in html format
+    #Uses the full unitable model - different to DetectionAndOcrTable1
+    def __init__(self):
+        self.unitableFullPredictor = UnitableFullPredictor()
+        self.detector = YOLO('foduucom/table-detection-and-extraction')
+        # set model parameters
+        self.detector.overrides['conf'] = 0.25 # NMS confidence threshold
+        self.detector.overrides['iou'] = 0.45  # NMS IoU threshold
+        self.detector.overrides['agnostic_nms'] = False  # NMS class-agnostic
+        self.detector.overrides['max_det'] = 1000  # maximum number of detections per image
+
+
+    
+    @staticmethod
+    def save_detection(detected_lines_images:List[ImageType], prefix = './res/test1/res_'):
+        i = 0
+        for img in detected_lines_images:
+            pilimg = Image.fromarray(img)
+            pilimg.save(prefix+str(i)+'.png')
+            i=i+1
+    """
+    Valid 'Boxes' object attributes and properties are:
+
+            Attributes:
+                boxes (torch.Tensor) or (numpy.ndarray): A tensor or numpy array containing the detection boxes,
+                    with shape (num_boxes, 6).
+                orig_shape (torch.Tensor) or (numpy.ndarray): Original image size, in the format (height, width).
+
+            Properties:
+                xyxy (torch.Tensor) or (numpy.ndarray): The boxes in xyxy format.
+                conf (torch.Tensor) or (numpy.ndarray): The confidence values of the boxes.
+                cls (torch.Tensor) or (numpy.ndarray): The class values of the boxes.
+                xywh (torch.Tensor) or (numpy.ndarray): The boxes in xywh format.
+                xyxyn (torch.Tensor) or (numpy.ndarray): The boxes in xyxy format normalized by original image size.
+                xywhn (torch.Tensor) or (numpy.ndarray): The boxes in xywh format normalized by original image size.
+    """
+    # Image is page image 
+    def predict(self,image:Image.Image,debugfolder_filename_page_name = None):
+        
+        results = self.detector.predict(image)
+
+        #Array of bboxes
+        bbxs = results[0].boxes.xyxy.int().tolist()
+        #Array of confidences 
+        conf = results[0].boxes.conf.float().tolist()
+        print(bbxs)
+        print(conf)
+    
+        #images_to_recognizer = cropImage(bxs, img)
+        img_to_save = draw_only_box(image, bbxs)
+        img_to_save.save(debugfolder_filename_page_name+"detectionBoxRes.png", quality=95)
+
+        # we need something to draw the detection 
+
+
+        cropped_tables =[]
+        for i in range (len(bbxs)):
+            # TODO: find the right confidence and padding values
+            if conf[i]< 0.65:
+                continue
+        
+            padded = [bbxs[i][0]-10,bbxs[i][1]-10,bbxs[i][2]+10,bbxs[i][3]+10]
+
+            cropped_table = image.convert("RGB").crop(padded)
+            cropped_table.save(debugfolder_filename_page_name +"yolo_cropped_table_"+str(i)+".png")
+            cropped_tables.append(cropped_table)
+
+        print("number of cropped tables found: "+str(len(cropped_tables)))
+        
+        # Step 1: Unitable 
+        #This take PIL Images as input 
+        if cropped_tables != []:
+           table_codes = self.unitableFullPredictor.predict(cropped_tables,debugfolder_filename_page_name)
+           return table_codes
+
+
+
+        
+
+
+
+        
+
+
+
+        

doctrfiles/__init__.py

@@ -0,0 +1,4 @@
+from .doctr_recognizer import DoctrTextRecognizer
+from .word_detector import Wordboxes,DoctrWordDetector
+
+__all__ = ['DoctrTextRecognizer','DoctrWordDetector','Wordboxes']

doctrfiles/doctr_recognizer.py

@@ -0,0 +1,183 @@
+import os
+from abc import ABC
+from pathlib import Path
+from typing import Any, List, Literal, Mapping, Optional, Tuple
+from zipfile import ZipFile
+import json
+from typing import Any, List, Literal, Mapping, Optional,Dict
+import uuid
+from doctr.models.preprocessor import PreProcessor
+from doctr.models.recognition.predictor import RecognitionPredictor  # pylint: disable=W0611
+from doctr.models.recognition.zoo import ARCHS, recognition
+import torch
+# Numpy image type
+import numpy.typing as npt
+from numpy import uint8
+ImageType = npt.NDArray[uint8]
+
+from utils import WordAnnotation,getlogger
+
+class DoctrTextRecognizer():
+
+    def __init__(
+        self,
+        architecture: str,
+        path_weights: str,
+        path_config_json: str = None,
+    ) -> None:
+        """
+        :param architecture: DocTR supports various text recognition models, e.g. "crnn_vgg16_bn",
+        "crnn_mobilenet_v3_small". The full list can be found here:
+        https://github.com/mindee/doctr/blob/main/doctr/models/recognition/zoo.py#L16.
+        :param path_weights: Path to the weights of the model
+        :param device: "cpu" or "cuda".
+        :param lib: "TF" or "PT" or None. If None, env variables USE_TENSORFLOW, USE_PYTORCH will be used.
+        :param path_config_json: Path to a json file containing the configuration of the model. Useful, if you have
+        a model trained on custom vocab.
+        """
+
+        self.architecture = architecture
+        self.path_weights = path_weights
+
+        self.name = self.get_name(self.path_weights, self.architecture)
+
+        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        self.path_config_json = path_config_json
+
+        self.built_model = self.build_model(self.architecture, self.path_config_json)
+        self.load_model(self.path_weights, self.built_model, self.device)
+        self.doctr_predictor = self.get_wrapped_model()
+
+    def predict(self, inputs: Dict[uuid.UUID, Tuple[ImageType,WordAnnotation]]) -> List[WordAnnotation]:
+        
+        """
+        Prediction on a batch of text lines
+
+        :param images: Dictionary where key is word's object id and the value is tupe of cropped image and word annotation
+        :return: A list of DetectionResult
+        """
+        if inputs:
+            
+
+            predictor =self.doctr_predictor
+            device = self.device
+            
+            word_uuids = list(inputs.keys())
+            cropped_images = [value[0] for value in inputs.values()]
+
+            raw_output = predictor(list(cropped_images))
+            det_results =[]
+            for uuid, output in zip(word_uuids, raw_output):
+                ann = inputs[uuid][1]
+                ann.text = output[0]
+                det_results.append(ann)
+            return det_results
+        return []
+    
+    def predict_for_tables(self, inputs: List[ImageType]) -> List[str]:
+
+        if inputs:
+
+            predictor =self.doctr_predictor
+            device = self.device
+            
+            raw_output = predictor(list(inputs))
+            det_results =[]
+            for output in raw_output:
+                det_results.append(output[0])
+            return det_results
+        return []
+
+    @staticmethod
+    def load_model(path_weights: str, doctr_predictor: Any, device: torch.device) -> None:
+        """Loading model weights
+        1. Load the State Dictionary:
+        state_dict = torch.load(path_weights, map_location=device) loads the state dictionary from the specified file path and maps it to the specified device.
+        2. Modify Keys in the State Dictionary:
+        The code prepends "model." to each key in the state dictionary. This is likely necessary to match the keys expected by the doctr_predictor model.
+        3. Load State Dictionary into Model:
+        doctr_predictor.load_state_dict(state_dict) loads the modified state dictionary into the model.
+        4. Move Model to Device:
+        doctr_predictor.to(device) moves the model to the specified device.
+        """
+        state_dict = torch.load(path_weights, map_location=device)
+        for key in list(state_dict.keys()):
+            state_dict["model." + key] = state_dict.pop(key)
+        doctr_predictor.load_state_dict(state_dict)
+        doctr_predictor.to(device)
+
+    @staticmethod
+    def build_model(architecture: str, path_config_json: Optional[str] = None) -> "RecognitionPredictor":
+        """Building the model
+        1. Specific keys (arch, url, task) are removed from custom_configs.
+            mean and std values are moved to recognition_configs.   
+        2. Creating model
+        Check Architecture Type:
+        Case 1 : 
+        If architecture is a string, it checks if it's in the predefined set of architectures (ARCHS).
+        If valid, it creates an instance of the model using the specified architecture and custom configurations.
+        Handle Custom Architecture Instances:
+        Case 2 : 
+        If architecture is not a string, it checks if it's an **instance** of one of the recognized model classes (e.g., recognition.CRNN, recognition.SAR, etc.).
+        If valid, it assigns the provided architecture to model.
+        Get Input Shape and Create RecognitionPredictor:
+
+        3. Retrieves the input_shape from the model's configuration.
+        4. Returns an instance of RecognitionPredictor initialized with a PreProcessor and the model.
+        """
+
+        # inspired and adapted from https://github.com/mindee/doctr/blob/main/doctr/models/recognition/zoo.py
+        custom_configs = {}
+        batch_size = 1024
+        recognition_configs = {}
+        if path_config_json:
+             with open(path_config_json, "r", encoding="utf-8") as f:
+                custom_configs = json.load(f)
+                custom_configs.pop("arch", None)
+                custom_configs.pop("url", None)
+                custom_configs.pop("task", None)
+                recognition_configs["mean"] = custom_configs.pop("mean")
+                recognition_configs["std"] = custom_configs.pop("std")
+                #batch_size = custom_configs.pop("batch_size")
+                recognition_configs["batch_size"] = batch_size
+
+        if isinstance(architecture, str):
+            if architecture not in ARCHS:
+                raise ValueError(f"unknown architecture '{architecture}'")
+
+            model = recognition.__dict__[architecture](pretrained=True, pretrained_backbone=True, **custom_configs)
+        else:
+            if not isinstance(
+                architecture,
+                (recognition.CRNN, recognition.SAR, recognition.MASTER, recognition.ViTSTR, recognition.PARSeq),
+            ):
+                raise ValueError(f"unknown architecture: {type(architecture)}")
+            model = architecture
+
+        input_shape = model.cfg["input_shape"][-2:]
+        """
+        (class) PreProcessor
+        Implements an abstract preprocessor object which performs casting, resizing, batching and normalization.
+
+        Args:
+            output_size: expected size of each page in format (H, W)
+            batch_size: the size of page batches
+            mean: mean value of the training distribution by channel
+            std: standard deviation of the training distribution by channel
+        """
+        return RecognitionPredictor(PreProcessor(input_shape, preserve_aspect_ratio=True, **recognition_configs), model)
+
+
+    def get_wrapped_model(self) -> Any:
+        """
+        Get the inner (wrapped) model.
+        """
+        doctr_predictor = self.build_model(self.architecture, self.path_config_json)
+        device_str = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        self.load_model(self.path_weights, doctr_predictor, device_str)
+        return doctr_predictor
+
+    @staticmethod
+    def get_name(path_weights: str, architecture: str) -> str:
+        """Returns the name of the model"""
+        return f"doctr_{architecture}" + "_".join(Path(path_weights).parts[-2:])

doctrfiles/models/config-multi2.json

@@ -0,0 +1,21 @@
+{
+  "mean": [
+    0.694,
+    0.695,
+    0.693
+  ],
+  "std": [
+    0.299,
+    0.296,
+    0.301
+  ],
+  "input_shape": [
+    3,
+    32,
+    128
+  ],
+  "vocab": "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~°£€¥¢฿äöüßÄÖÜẞàâéèêëîïôùûçÀÂÉÈÊËÎÏÔÙÛǧ",
+  "url": "https://doctr-static.mindee.com/models?id=v0.3.1/crnn_vgg16_bn-9762b0b0.pt&src=0",
+  "arch": "crnn_vgg16_bn",
+  "task": "recognition"
+}

doctrfiles/models/db_mobilenet_v3_large-81e9b152.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81e9b152c11e9681f5eb4a2ec72e5f5d67df8ab860a846e1004756badfa5d37a
+size 16987510

doctrfiles/models/db_resnet34-cb6aed9e.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb6aed9e4252c8a92d381de1b15e1e75461f7a125a4262ef16768a4b9f797347
+size 89991042

doctrfiles/models/db_resnet50-79bd7d70.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79bd7d702506703b89cf11afa42d12aebf5cf25c3618e6ffd5f85772240ca483
+size 102021912

doctrfiles/models/db_resnet50_config.json

@@ -0,0 +1,20 @@
+{
+"mean": [
+   0.798,
+   0.785,
+   0.772
+],
+"std": [
+   0.264,
+   0.2749,
+   0.287
+],
+"input_shape": [
+   3,
+   1024,
+   1024
+],
+"url": "https://doctr-static.mindee.com/models?id=v0.7.0/parseq-56125471.pt&src=0",
+"arch": "db_resnet50",
+"task": "detection"
+}

doctrfiles/models/doctr-multilingual-parseq.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5b1d3f3b9d8ab994e94c671c47828b9a4079f20b4288eb5f0ba3c6dacf6c237
+size 47872130

doctrfiles/models/master-fde31e4a.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fde31e4a9612670af83daf4b730dd9c56216806589546b09290abc347ca3a49d
+size 243889428

doctrfiles/models/master.json

@@ -0,0 +1,21 @@
+{
+  "mean": [
+    0.694,
+    0.695,
+    0.693
+  ],
+  "std": [
+    0.299,
+    0.296,
+    0.301
+  ],
+  "input_shape": [
+    3,
+    32,
+    128
+  ],
+  "vocab": "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~°£€¥¢฿àâéèêëîïôùûüçÀÂÉÈÊËÎÏÔÙÛÜÇ",
+  "url": null,
+  "arch": "master",
+  "task": "recognition"
+}

doctrfiles/models/multi2.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bdc3c6922cad527714504b84a9d0efaa6b679d8ca8050a003611076eb514757
+size 63310142

doctrfiles/models/multilingual-parseq-config.json

@@ -0,0 +1,21 @@
+{
+  "mean": [
+    0.694,
+    0.695,
+    0.693
+  ],
+  "std": [
+    0.299,
+    0.296,
+    0.301
+  ],
+  "input_shape": [
+    3,
+    32,
+    128
+  ],
+  "vocab": "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~°£€¥¢฿àâéèêëîïôùûüçÀÂÉÈÊËÎÏÔÙÛÜÇáãíóõúÁÃÍÓÕÚñÑ¡¿äößÄÖẞčďěňřšťůýžČĎĚŇŘŠŤŮÝŽąćęłńśźżĄĆĘŁŃŚŹŻìòÌÒæøåÆØŧ",
+  "url": "https://doctr-static.mindee.com/models?id=v0.7.0/parseq-56125471.pt&src=0",
+  "arch": "parseq",
+  "task": "recognition"
+}

doctrfiles/word_detector.py

@@ -0,0 +1,282 @@
+
+import os
+from abc import ABC
+from pathlib import Path
+from typing import Any, List, Literal, Mapping, Optional, Tuple, Union, Dict, Type, Sequence
+import json
+import logging
+import torch
+from doctr.models.preprocessor import PreProcessor
+from doctr.models.detection.predictor import DetectionPredictor  # pylint: disable=W0611
+from doctr.models.detection.zoo import detection_predictor,detection
+
+import numpy.typing as npt
+import numpy as np
+from numpy import uint8
+ImageType = npt.NDArray[uint8]
+
+
+from utils import Annotation,getlogger,group_words_into_lines
+
+ARCHS = [
+        "db_resnet34",
+        "db_resnet50",
+        "db_mobilenet_v3_large",
+        "linknet_resnet18",
+        "linknet_resnet34",
+        "linknet_resnet50",
+        "fast_tiny",
+        "fast_small",
+        "fast_base",
+    ]
+class Wordboxes:
+    def __init__(self,score, box):
+        self.box = box 
+        self.score = score 
+
+class DoctrWordDetector():
+    """
+    A deepdoctection wrapper of DocTr text line detector. We model text line detection as ObjectDetector
+    and assume to use this detector in a ImageLayoutService.
+    DocTr supports several text line detection implementations but provides only a subset of pre-trained models.
+    The most usable one for document OCR for which a pre-trained model exists is DBNet as described in “Real-time Scene
+    Text Detection with Differentiable Binarization”, with a ResNet-50 backbone. This model can be used in either
+    Tensorflow or PyTorch.
+    Some other pre-trained models exist that have not been registered in `ModelCatalog`. Please check the DocTr library
+    and organize the download of the pre-trained model by yourself.
+
+    **Example:**
+
+                 path_weights_tl = ModelDownloadManager.maybe_download_weights_and_configs("doctr/db_resnet50/pt
+                 /db_resnet50-ac60cadc.pt")
+                 # Use "doctr/db_resnet50/tf/db_resnet50-adcafc63.zip" for Tensorflow
+
+                 categories = ModelCatalog.get_profile("doctr/db_resnet50/pt/db_resnet50-ac60cadc.pt").categories
+                 det = DoctrTextlineDetector("db_resnet50",path_weights_tl,categories,"cpu")
+                 layout = ImageLayoutService(det,to_image=True, crop_image=True)
+
+                 path_weights_tr = dd.ModelDownloadManager.maybe_download_weights_and_configs("doctr/crnn_vgg16_bn
+                 /pt/crnn_vgg16_bn-9762b0b0.pt")
+                 rec = DoctrTextRecognizer("crnn_vgg16_bn", path_weights_tr, "cpu")
+                 text = TextExtractionService(rec, extract_from_roi="word")
+
+                 analyzer = DoctectionPipe(pipeline_component_list=[layout,text])
+
+                 path = "/path/to/image_dir"
+                 df = analyzer.analyze(path = path)
+
+                 for dp in df:
+                     ...
+    """
+
+    def __init__(
+        self,
+        architecture: str,
+        path_weights: str,
+        path_config_json:str
+    ) -> None:
+        """
+        :param architecture: DocTR supports various text line detection models, e.g. "db_resnet50",
+        "db_mobilenet_v3_large". The full list can be found here:
+        https://github.com/mindee/doctr/blob/main/doctr/models/detection/zoo.py#L20
+        :param path_weights: Path to the weights of the model
+        :param categories: A dict with the model output label and value
+        :param device: "cpu" or "cuda" or any tf.device or torch.device. The device must be compatible with the dll
+        :param lib: "TF" or "PT" or None. If None, env variables USE_TENSORFLOW, USE_PYTORCH will be used.
+        """
+        self.architecture = architecture
+        self.path_weights = path_weights
+        self.path_config_json =path_config_json
+
+        # Ensure the correct device is chosen (either CPU or CUDA if available)
+        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+        # Initialize the model with the given architecture and path to weights
+        self.doctr_predictor = self.get_wrapped_model()
+
+    """
+    Two static method so that they can be called without creating an instance of the class
+    Also, they don't require any instance specific data 
+    """
+
+    def get_wrapped_model(
+        self
+    ) -> Any:
+        """
+        Get the inner (wrapped) model.
+
+        :param architecture: DocTR supports various text line detection models, e.g. "db_resnet50",
+        "db_mobilenet_v3_large". The full list can be found here:
+        https://github.com/mindee/doctr/blob/main/doctr/models/detection/zoo.py#L20
+        :param path_weights: Path to the weights of the model
+
+        :return: Inner model which is a "nn.Module" in PyTorch or a "tf.keras.Model" in Tensorflow
+        """
+
+        """
+        (function) detection_predictor: ((arch: Any = "db_resnet50", pretrained: bool = False, assume_straight_pages: bool = True, **kwargs: Any) -> DetectionPredictor) 
+        """
+        #doctr_predictor = detection_predictor(arch=architecture, pretrained=False, pretrained_backbone=False)
+        #doctr_predictor = detection_predictor(arch=architecture, pretrained=False)
+
+        doctr_predictor = self.build_model(self.architecture, self.path_config_json)
+       
+        self.load_model(self.path_weights, doctr_predictor, self.device)
+        return doctr_predictor
+    @staticmethod
+    def build_model(arch: str, pretrained = False,assume_straight_pages=True, path_config_json: Optional[str] = None) -> "DetectionPredictor":
+        """Building the model
+        1. Specific keys (arch, url, task) are removed from custom_configs.
+            mean and std values are moved to recognition_configs.   
+        2. Creating model
+        Check Architecture Type:
+        Case 1 : 
+        If architecture is a string, it checks if it's in the predefined set of architectures (ARCHS).
+        If valid, it creates an instance of the model using the specified architecture and custom configurations.
+        Handle Custom Architecture Instances:
+        Case 2 : 
+        If architecture is not a string, it checks if it's an **instance** of one of the recognized model classes (e.g., recognition.CRNN, recognition.SAR, etc.).
+        If valid, it assigns the provided architecture to model.
+        Get Input Shape and Create RecognitionPredictor:
+
+        3. Retrieves the input_shape from the model's configuration.
+        4. Returns an instance of RecognitionPredictor initialized with a PreProcessor and the model.
+        """
+
+        custom_configs = {}
+        batch_size = 4
+        detection_configs = {}
+        if path_config_json:
+             with open(path_config_json, "r", encoding="utf-8") as f:
+                custom_configs = json.load(f)
+                custom_configs.pop("arch", None)
+                custom_configs.pop("url", None)
+                custom_configs.pop("task", None)
+                detection_configs["mean"] = custom_configs.pop("mean")
+                detection_configs["std"] = custom_configs.pop("std")
+                #batch_size = custom_configs.pop("batch_size")
+                detection_configs["batch_size"] = batch_size
+        if isinstance(arch, str):
+            if arch not in ARCHS:
+                raise ValueError(f"unknown architecture '{arch}'")
+
+            model = detection.__dict__[arch](
+                pretrained=pretrained,
+                assume_straight_pages=assume_straight_pages
+            )
+
+        else:
+            if not isinstance(arch, (detection.DBNet, detection.LinkNet, detection.FAST)):
+                raise ValueError(f"unknown architecture: {type(arch)}")
+
+            model = arch
+            model.assume_straight_pages = assume_straight_pages
+
+        input_shape = model.cfg["input_shape"][-2:]
+
+        predictor = DetectionPredictor(
+            PreProcessor(input_shape, batch_size=batch_size,**detection_configs),
+            model
+        )
+        return predictor
+
+    @staticmethod
+    def load_model(path_weights: str, doctr_predictor: Any, device: torch.device) -> None:
+        """Loading model weights
+        1. Load the State Dictionary:
+        state_dict = torch.load(path_weights, map_location=device) loads the state dictionary from the specified file path and maps it to the specified device.
+        2. Modify Keys in the State Dictionary:
+        The code prepends "model." to each key in the state dictionary. This is likely necessary to match the keys expected by the doctr_predictor model.
+        3. Load State Dictionary into Model:
+        doctr_predictor.load_state_dict(state_dict) loads the modified state dictionary into the model.
+        4. Move Model to Device:
+        doctr_predictor.to(device) moves the model to the specified device.
+        """
+        state_dict = torch.load(path_weights, map_location=device)
+        for key in list(state_dict.keys()):
+            state_dict["model." + key] = state_dict.pop(key)
+        doctr_predictor.load_state_dict(state_dict)
+        doctr_predictor.to(device)
+
+    
+    def predict(self, np_img: ImageType,sort_vertical = False) -> List[Wordboxes]:
+        """
+        Prediction per image.
+
+        :param np_img: image as numpy array
+        
+        :return: A list of DetectionResult
+        """
+        
+        raw_output =self.doctr_predictor([np_img])
+        height, width = np_img.shape[:2]
+
+        """
+        raw_output is arrary of dictionary with just one key "words"
+        1-4th element : coordinates You take first 4 elements in this array by doing box[:4]
+        5th element - score 
+        But those are 4 point and we need 4X2 
+        type(raw_output[0]["words"]) are numpy arrary 
+        Okay hypothesis :xmin, ymin, xmax, ymax
+        Points should be ordered in this order :left_lower, right_lower, right_upper, left_upper
+        """
+        
+        logger = getlogger("array")
+        # Check if the logger has any handlers
+        if (logger.hasHandlers()):
+            logger.handlers.clear()
+
+        # Create a handler
+        handler = logging.StreamHandler()
+
+        # Create a formatter and add it to the handler
+        formatter = logging.Formatter('%(levelname)s:%(message)s')
+        handler.setFormatter(formatter)
+
+        # Add the handler to the logger
+        logger.addHandler(handler)
+        #logger.info(raw_output[0]["words"])
+        
+        #array is numpy array of shape (n,5) where n is number of words and 5 is size of each element(array) with  coordinate(xmin,ymin,xmax,ymax) + score 
+        
+        array = raw_output[0]["words"]
+        if not sort_vertical:
+            #Only When input has one line 
+            sorted_array = array[array[:, 0].argsort()]
+        else:
+            #When input can have multiple lines
+            sorted_array = group_words_into_lines(array)
+        #logger.info(sorted_array)
+
+
+        detection_results = []
+        for box in sorted_array:
+            xmin, ymin, xmax, ymax = box[:4]
+            xmin = xmin*width
+            ymin = ymin*height
+            xmax = xmax*width
+            ymax = ymax*height
+            newb = np.array([
+                    [xmin, ymin],
+                    [xmax, ymin],
+                    [xmax, ymax],
+                    [xmin, ymax]
+                ], dtype=np.float32)
+            assert newb.shape == (4, 2), f"Points array must be of shape (4, 2), but got {box.shape}"
+            assert newb.dtype == np.float32, f"Points array must be of dtype float32, but got {box.dtype}"
+
+            w = Wordboxes(
+                score=box[4],
+                box = newb
+            )
+            
+            detection_results.append(w)
+
+        return detection_results
+
+    
+
+
+    
+
+