Duplicate from tianchiguaixia/ocr

.gitattributes

@@ -0,0 +1,34 @@
+*.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
+*.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

README.md

@@ -0,0 +1,13 @@
+---
+title: Ocr
+emoji: 🐠
+colorFrom: green
+colorTo: green
+sdk: streamlit
+sdk_version: 1.21.0
+app_file: app.py
+pinned: false
+duplicated_from: tianchiguaixia/ocr
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,67 @@
+# -*- coding: utf-8 -*-
+# time: 2022/10/17 11:22
+# file: AI-医学图片OCR.py
+
+
+import streamlit as st
+
+from ocr.ocr import detect, recognize
+from ocr.utils import bytes_to_numpy
+import pandas as pd
+
+import os
+import cv2
+from paddleocr import PPStructure, draw_structure_result, save_structure_res
+
+
+st.title("AI-医学图片OCR")
+
+
+
+def convert_df(df):
+    # IMPORTANT: Cache the conversion to prevent computation on every rerun
+    return df.to_csv().encode("gbk")
+
+
+# 上传图片
+uploaded_file = st.sidebar.file_uploader(
+    '请选择一张图片', type=['png', 'jpg', 'jpeg'])
+print('uploaded_file:', uploaded_file)
+table_engine = PPStructure(show_log=True)
+if uploaded_file is not None:
+    # To read file as bytes:
+    # content = cv2.imread(uploaded_file)
+    # st.write(content)
+    bytes_data = uploaded_file.getvalue()
+    # 转换格式
+    img = bytes_to_numpy(bytes_data, channels='RGB')
+    option_task = st.sidebar.radio('请选择要执行的任务', ('查看原图', '文本检测'))
+    if option_task == '查看原图':
+        st.image(img, caption='原图')
+    elif option_task == '文本检测':
+        im_show = detect(img)
+        st.image(im_show, caption='文本检测后的图片')
+
+    base_path = "streamlit_data"
+
+    path = os.path.exists(base_path + "/" + uploaded_file.name.split('.')[0])
+
+    if st.button('✨ 启动!'):
+        local_path = base_path + "/" + uploaded_file.name.split('.')[0]
+        result = table_engine(img)
+        save_structure_res(result, base_path, uploaded_file.name.split('.')[0])
+        with st.container():
+            with st.expander(label="json结果展示", expanded=False):
+                st.write(result)
+            for i in os.listdir(local_path):
+                if ".xlsx" in i:
+                    df = pd.read_excel(os.path.join(local_path, i))
+                    df = df.fillna("")
+                    st.write(df)
+                    csv = convert_df(df)
+                    st.download_button(
+                        label="Download data as csv",
+                        data=csv,
+                        file_name='large_df.csv',
+                        mime='text/csv',
+                    )

ocr/__init__.py

@@ -0,0 +1,5 @@
+# -*- coding: utf-8 -*-
+# time: 2022/10/17 13:03
+# file: __init__.py.py
+# author: fengchenggang
+# email: chenggang.feng@ashermed.com

ocr/ocr.py

@@ -0,0 +1,45 @@
+# -*- coding: utf-8 -*-
+# time: 2022/10/17 13:04
+# file: ocr.py
+
+
+from paddleocr import PaddleOCR
+from ocr_utils import draw_ocr, draw_ocr_box_txt
+
+ocr = PaddleOCR(lang='ch',use_angle_cls=True)
+
+
+def detect(image):
+    """
+    文本检测
+    args:
+        image(array): numpy格式的图片 'RGB'
+    return(array):
+        检测后的图片 numpy格式 'RGB'
+    """
+    result = ocr.ocr(image, rec=False)
+    im_show = draw_ocr(image, result)
+
+    return im_show
+
+
+def recognize(image, output_mode=0):
+    """
+    文本识别
+    args:
+        image(array): numpy格式的图片 'RGB'
+        output_mode(int): 图片输出模式 [0|1]
+    return(array):
+        识别后的图片 numpy格式 'RGB'
+    """
+    result = ocr.ocr(image)
+    boxes = [line[0] for line in result]
+    txts = [line[1][0] for line in result]
+    scores = [line[1][1] for line in result]
+    if output_mode == 0:
+        im_show = draw_ocr_box_txt(image, boxes, txts, scores)
+    elif output_mode == 1:
+        im_show = draw_ocr(image, boxes, txts, scores)
+
+    return im_show
+

ocr/utils.py

@@ -0,0 +1,40 @@
+# -*- coding: utf-8 -*-
+# time: 2022/10/17 13:04
+# file: utils.py
+
+
+import cv2
+import numpy as np
+
+
+def bytes_to_numpy(image_bytes, channels='BGR'):
+    """
+    图片格式转换 bytes -> numpy
+    args:
+        image_bytes(str): 图片的字节流
+        channels(str): 图片的格式 ['BGR'|'RGB']
+    return(array):
+        转换后的图片
+    """
+    _image_np = np.frombuffer(image_bytes, dtype=np.uint8)
+    image_np = cv2.imdecode(_image_np, cv2.IMREAD_COLOR)
+    if channels == 'BGR':
+        return image_np
+    elif channels == 'RGB':
+        image_np = cv2.cvtColor(image_np, cv2.COLOR_BGR2RGB)
+        return image_np
+
+
+def numpy_to_pic(image_path, image_np, channels='BGR'):
+    """
+    保存图片
+    args:
+        image_path(str): 图片路径
+        image_np(array): numpy格式的图片数据
+        channels(str): 图片的格式 ['BGR'|'RGB']
+    """
+    if channels == 'BGR':
+        cv2.imwrite(image_path, image_np)
+    elif channels == 'RGB':
+        image_np = cv2.cvtColor(image_np, cv2.COLOR_RGB2BGR)
+        cv2.imwrite(image_path, image_np)

ocr_utils.py

@@ -0,0 +1,281 @@
+# -*- coding: utf-8 -*-
+# time: 2022/10/17 13:25
+# file: ocr_utils.py
+
+import cv2
+import math
+import numpy as np
+
+from PIL import Image, ImageDraw, ImageFont
+
+
+def resize_img(img, input_size=600):
+    """
+    resize img and limit the longest side of the image to input_size
+    """
+    img = np.array(img)
+    im_shape = img.shape
+    im_size_max = np.max(im_shape[0:2])
+    im_scale = float(input_size) / float(im_size_max)
+    img = cv2.resize(img, None, None, fx=im_scale, fy=im_scale)
+    return img
+
+
+def draw_ocr(
+    image,
+    boxes,
+    txts=None,
+    scores=None,
+    drop_score=0.5,
+    font_path="./fonts/font.ttf"
+):
+    """
+    Visualize the results of OCR detection and recognition
+    args:
+        image(Image|array): RGB image
+        boxes(list): boxes with shape(N, 4, 2)
+        txts(list): the texts
+        scores(list): txxs corresponding scores
+        drop_score(float): only scores greater than drop_threshold will be visualized
+        font_path: the path of font which is used to draw text
+    return(array):
+        the visualized img
+    """
+    if scores is None:
+        scores = [1] * len(boxes)
+    box_num = len(boxes)
+    for i in range(box_num):
+        if scores is not None and (scores[i] < drop_score or math.isnan(scores[i])):
+            continue
+        box = np.reshape(np.array(boxes[i]), [-1, 1, 2]).astype(np.int64)
+        image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2)
+    if txts is not None:
+        img = np.array(resize_img(image, input_size=600))
+        txt_img = text_visual(
+            txts,
+            scores,
+            img_h=img.shape[0],
+            img_w=600,
+            threshold=drop_score,
+            font_path=font_path
+        )
+        img = np.concatenate([np.array(img), np.array(txt_img)], axis=1)
+        return img
+    return image
+
+
+def draw_ocr_box_txt(
+    image,
+    boxes,
+    txts,
+    scores=None,
+    drop_score=0.5,
+    font_path="./fonts/font.ttf"
+):
+    image = Image.fromarray(image)
+    h, w = image.height, image.width
+    img_left = image.copy()
+    img_right = Image.new('RGB', (w, h), (255, 255, 255))
+
+    import random
+
+    random.seed(0)
+    draw_left = ImageDraw.Draw(img_left)
+    draw_right = ImageDraw.Draw(img_right)
+    for idx, (box, txt) in enumerate(zip(boxes, txts)):
+        if scores is not None and scores[idx] < drop_score:
+            continue
+        color = (random.randint(0, 255), random.randint(0, 255), random.randint(0, 255))
+        draw_left.polygon(
+            [
+                box[0][0], box[0][1], box[1][0], box[1][1], box[2][0],
+                box[2][1], box[3][0], box[3][1]
+            ],
+            fill=color)
+        draw_right.polygon(
+            [
+                box[0][0], box[0][1], box[1][0], box[1][1], box[2][0],
+                box[2][1], box[3][0], box[3][1]
+            ],
+            outline=color)
+        box_height = math.sqrt((box[0][0] - box[3][0])**2 + (box[0][1] - box[3][
+            1])**2)
+        box_width = math.sqrt((box[0][0] - box[1][0])**2 + (box[0][1] - box[1][
+            1])**2)
+        if box_height > 2 * box_width:
+            font_size = max(int(box_width * 0.9), 10)
+            font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
+            cur_y = box[0][1]
+            for c in txt:
+                char_size = font.getsize(c)
+                draw_right.text((box[0][0] + 3, cur_y), c, fill=(0, 0, 0), font=font)
+                cur_y += char_size[1]
+        else:
+            font_size = max(int(box_height * 0.8), 10)
+            font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
+            draw_right.text([box[0][0], box[0][1]], txt, fill=(0, 0, 0), font=font)
+    img_left = Image.blend(image, img_left, 0.5)
+    img_show = Image.new('RGB', (w * 2, h), (255, 255, 255))
+    img_show.paste(img_left, (0, 0, w, h))
+    img_show.paste(img_right, (w, 0, w * 2, h))
+    return np.array(img_show)
+
+
+def str_count(s):
+    """
+    Count the number of Chinese characters,
+    a single English character and a single number
+    equal to half the length of Chinese characters.
+    args:
+        s(string): the input of string
+    return(int):
+        the number of Chinese characters
+    """
+    import string
+    count_zh = count_pu = 0
+    s_len = len(s)
+    en_dg_count = 0
+    for c in s:
+        if c in string.ascii_letters or c.isdigit() or c.isspace():
+            en_dg_count += 1
+        elif c.isalpha():
+            count_zh += 1
+        else:
+            count_pu += 1
+    return s_len - math.ceil(en_dg_count / 2)
+
+
+def text_visual(
+    texts,
+    scores,
+    img_h=400,
+    img_w=600,
+    threshold=0.,
+    font_path="./fonts/font.ttf"
+):
+    """
+    create new blank img and draw txt on it
+    args:
+        texts(list): the text will be draw
+        scores(list|None): corresponding score of each txt
+        img_h(int): the height of blank img
+        img_w(int): the width of blank img
+        font_path: the path of font which is used to draw text
+    return(array):
+    """
+    if scores is not None:
+        assert len(texts) == len(scores), "The number of txts and corresponding scores must match"
+
+    def create_blank_img():
+        blank_img = np.ones(shape=[img_h, img_w], dtype=np.int8) * 255
+        blank_img[:, img_w - 1:] = 0
+        blank_img = Image.fromarray(blank_img).convert("RGB")
+        draw_txt = ImageDraw.Draw(blank_img)
+        return blank_img, draw_txt
+
+    blank_img, draw_txt = create_blank_img()
+
+    font_size = 20
+    txt_color = (0, 0, 0)
+    font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
+
+    gap = font_size + 5
+    txt_img_list = []
+    count, index = 1, 0
+    for idx, txt in enumerate(texts):
+        index += 1
+        if scores[idx] < threshold or math.isnan(scores[idx]):
+            index -= 1
+            continue
+        first_line = True
+        while str_count(txt) >= img_w // font_size - 4:
+            tmp = txt
+            txt = tmp[:img_w // font_size - 4]
+            if first_line:
+                new_txt = str(index) + ': ' + txt
+                first_line = False
+            else:
+                new_txt = '    ' + txt
+            draw_txt.text((0, gap * count), new_txt, txt_color, font=font)
+            txt = tmp[img_w // font_size - 4:]
+            if count >= img_h // gap - 1:
+                txt_img_list.append(np.array(blank_img))
+                blank_img, draw_txt = create_blank_img()
+                count = 0
+            count += 1
+        if first_line:
+            new_txt = str(index) + ': ' + txt + '   ' + '%.3f' % (scores[idx])
+        else:
+            new_txt = "  " + txt + "  " + '%.3f' % (scores[idx])
+        draw_txt.text((0, gap * count), new_txt, txt_color, font=font)
+        # whether add new blank img or not
+        if count >= img_h // gap - 1 and idx + 1 < len(texts):
+            txt_img_list.append(np.array(blank_img))
+            blank_img, draw_txt = create_blank_img()
+            count = 0
+        count += 1
+    txt_img_list.append(np.array(blank_img))
+    if len(txt_img_list) == 1:
+        blank_img = np.array(txt_img_list[0])
+    else:
+        blank_img = np.concatenate(txt_img_list, axis=1)
+    return np.array(blank_img)
+
+
+def base64_to_cv2(b64str):
+    import base64
+    data = base64.b64decode(b64str.encode('utf8'))
+    data = np.fromstring(data, np.uint8)
+    data = cv2.imdecode(data, cv2.IMREAD_COLOR)
+    return data
+
+
+def draw_boxes(image, boxes, scores=None, drop_score=0.5):
+    if scores is None:
+        scores = [1] * len(boxes)
+    for (box, score) in zip(boxes, scores):
+        if score < drop_score:
+            continue
+        box = np.reshape(np.array(box), [-1, 1, 2]).astype(np.int64)
+        image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2)
+    return image
+
+
+def get_rotate_crop_image(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
+
+
+if __name__ == '__main__':
+    pass

requirements.txt

@@ -0,0 +1,9 @@
+paddleocr~=2.6
+paddlepaddle
+opencv-python==4.6.0.66
+opencv-contrib-python==4.6.0.66
+opencv-python-headless
+numpy~=1.23.4
+pillow~=8.0.1
+streamlit~=1.12.2
+pandas~=1.5.0