app.py

from transformers import TrOCRProcessor, VisionEncoderDecoderModel
from PIL import Image
import requests
import warnings
from skimage.io import imread
from skimage.color import rgb2gray
import matplotlib.pyplot as plt
from skimage.filters import sobel
import numpy as np
from heapq import *
import gradio as gr
from skimage.filters import threshold_otsu
from skimage.util import invert
import imageio
from matplotlib.dates import SU
from regex import F
from sklearn.feature_extraction.text import TfidfVectorizer
from sentence_transformers import SentenceTransformer, util
from sklearn.metrics.pairwise import cosine_similarity
import spacy
import pandas as pd
from tqdm import tqdm
import textdistance
from spacy.lang.en.stop_words import STOP_WORDS
#import psycopg2
import os
from tensorflow.keras.applications.resnet50 import ResNet50,preprocess_input, decode_predictions
from tensorflow.keras.preprocessing import image
from sklearn.feature_extraction.text import TfidfVectorizer




processor = TrOCRProcessor.from_pretrained('microsoft/trocr-base-handwritten')
model = VisionEncoderDecoderModel.from_pretrained('microsoft/trocr-base-handwritten')
plt.switch_backend('Agg')
def horizontal_projections(sobel_image):
    return np.sum(sobel_image, axis=1)  
            
            
def find_peak_regions(hpp, divider=4):
    threshold = (np.max(hpp)-np.min(hpp))/divider
    peaks = []
    
    for i, hppv in enumerate(hpp):
        if hppv < threshold:
            peaks.append([i, hppv])
    return peaks

def heuristic(a, b):
    return (b[0] - a[0]) ** 2 + (b[1] - a[1]) ** 2

def get_hpp_walking_regions(peaks_index):
    hpp_clusters = []
    cluster = []
    for index, value in enumerate(peaks_index):
        cluster.append(value)

        if index < len(peaks_index)-1 and peaks_index[index+1] - value > 1:
            hpp_clusters.append(cluster)
            cluster = []

        #get the last cluster
        if index == len(peaks_index)-1:
            hpp_clusters.append(cluster)
            cluster = []
            
    return hpp_clusters

def astar(array, start, goal):

        neighbors = [(0,1),(0,-1),(1,0),(-1,0),(1,1),(1,-1),(-1,1),(-1,-1)]
        close_set = set()
        came_from = {}
        gscore = {start:0}
        fscore = {start:heuristic(start, goal)}
        oheap = []

        heappush(oheap, (fscore[start], start))
        
        while oheap:

            current = heappop(oheap)[1]

            if current == goal:
                data = []
                while current in came_from:
                    data.append(current)
                    current = came_from[current]
                return data

            close_set.add(current)
            for i, j in neighbors:
                neighbor = current[0] + i, current[1] + j            
                tentative_g_score = gscore[current] + heuristic(current, neighbor)
                if 0 <= neighbor[0] < array.shape[0]:
                    if 0 <= neighbor[1] < array.shape[1]:                
                        if array[neighbor[0]][neighbor[1]] == 1:
                            continue
                    else:
                        # array bound y walls
                        continue
                else:
                    # array bound x walls
                    continue
                    
                if neighbor in close_set and tentative_g_score >= gscore.get(neighbor, 0):
                    continue
                    
                if  tentative_g_score < gscore.get(neighbor, 0) or neighbor not in [i[1]for i in oheap]:
                    came_from[neighbor] = current
                    gscore[neighbor] = tentative_g_score
                    fscore[neighbor] = tentative_g_score + heuristic(neighbor, goal)
                    heappush(oheap, (fscore[neighbor], neighbor))
                    
        return []

def get_binary(img):
        mean = np.mean(img)
        if mean == 0.0 or mean == 1.0:
            return img

        thresh = threshold_otsu(img)
        binary = img <= thresh
        binary = binary*1
        return binary

def path_exists(window_image):
    #very basic check first then proceed to A* check
    if 0 in horizontal_projections(window_image):
        return True
    
    padded_window = np.zeros((window_image.shape[0],1))
    world_map = np.hstack((padded_window, np.hstack((window_image,padded_window)) ) )
    path = np.array(astar(world_map, (int(world_map.shape[0]/2), 0), (int(world_map.shape[0]/2), world_map.shape[1])))
    if len(path) > 0:
        return True
    
    return False

def get_road_block_regions(nmap):
    road_blocks = []
    needtobreak = False
    
    for col in range(nmap.shape[1]):
        start = col
        end = col+20
        if end > nmap.shape[1]-1:
            end = nmap.shape[1]-1
            needtobreak = True

        if path_exists(nmap[:, start:end]) == False:
            road_blocks.append(col)

        if needtobreak == True:
            break
            
    return road_blocks

def group_the_road_blocks(road_blocks):
    #group the road blocks
    road_blocks_cluster_groups = []
    road_blocks_cluster = []
    size = len(road_blocks)
    for index, value in enumerate(road_blocks):
        road_blocks_cluster.append(value)
        if index < size-1 and (road_blocks[index+1] - road_blocks[index]) > 1:
            road_blocks_cluster_groups.append([road_blocks_cluster[0], road_blocks_cluster[len(road_blocks_cluster)-1]])
            road_blocks_cluster = []

        if index == size-1 and len(road_blocks_cluster) > 0:
            road_blocks_cluster_groups.append([road_blocks_cluster[0], road_blocks_cluster[len(road_blocks_cluster)-1]])
            road_blocks_cluster = []

    return road_blocks_cluster_groups

def extract_line_from_image(image, lower_line, upper_line):
    lower_boundary = np.min(lower_line[:, 0])
    upper_boundary = np.min(upper_line[:, 0])
    img_copy = np.copy(image)
    r, c = img_copy.shape
    for index in range(c-1):
        img_copy[0:lower_line[index, 0], index] = 0
        img_copy[upper_line[index, 0]:r, index] = 0
    
    return img_copy[lower_boundary:upper_boundary, :]

def extract(image):
    img = rgb2gray(image)

    #img = rgb2gray(imread("Penwritten_2048x.jpeg"))
    #img = rgb2gray(imread("test.jpg"))
    #img = rgb2gray(imread(""))




    sobel_image = sobel(img)
    hpp = horizontal_projections(sobel_image)


    warnings.filterwarnings("ignore")
    #find the midway where we can make a threshold and extract the peaks regions
    #divider parameter value is used to threshold the peak values from non peak values.


    peaks = find_peak_regions(hpp)

    peaks_index = np.array(peaks)[:,0].astype(int)
    #print(peaks_index.shape)
    segmented_img = np.copy(img)
    r= segmented_img.shape
    for ri in range(r[0]):
        if ri in peaks_index:
            segmented_img[ri, :] = 0

    #group the peaks into walking windows


    hpp_clusters = get_hpp_walking_regions(peaks_index)
    #a star path planning algorithm 
    



    


    #Scan the paths to see if there are any blockers.
    

    

    binary_image = get_binary(img)

    for cluster_of_interest in hpp_clusters:
        nmap = binary_image[cluster_of_interest[0]:cluster_of_interest[len(cluster_of_interest)-1],:]
        road_blocks = get_road_block_regions(nmap)
        road_blocks_cluster_groups = group_the_road_blocks(road_blocks)
        #create the doorways
        for index, road_blocks in enumerate(road_blocks_cluster_groups):
            window_image = nmap[:, road_blocks[0]: road_blocks[1]+10]
            binary_image[cluster_of_interest[0]:cluster_of_interest[len(cluster_of_interest)-1],:][:, road_blocks[0]: road_blocks[1]+10][int(window_image.shape[0]/2),:] *= 0

    #now that everything is cleaner, its time to segment all the lines using the A* algorithm
    line_segments = []
    #print(len(hpp_clusters))
    #print(hpp_clusters)
    for i, cluster_of_interest in enumerate(hpp_clusters):
        nmap = binary_image[cluster_of_interest[0]:cluster_of_interest[len(cluster_of_interest)-1],:]
        path = np.array(astar(nmap, (int(nmap.shape[0]/2), 0), (int(nmap.shape[0]/2),nmap.shape[1]-1)))
        #print(path.shape)
        if path.shape[0]!=0:
            #break
            offset_from_top = cluster_of_interest[0]
            #print(offset_from_top)
            path[:,0] += offset_from_top
            #print(path)
            line_segments.append(path)
            #print(i)

    cluster_of_interest = hpp_clusters[1]
    offset_from_top = cluster_of_interest[0]
    nmap = binary_image[cluster_of_interest[0]:cluster_of_interest[len(cluster_of_interest)-1],:]
    #plt.figure(figsize=(20,20))
    #plt.imshow(invert(nmap), cmap="gray")

    path = np.array(astar(nmap, (int(nmap.shape[0]/2), 0), (int(nmap.shape[0]/2),nmap.shape[1]-1)))
    #plt.plot(path[:,1], path[:,0])

    offset_from_top = cluster_of_interest[0]



    ## add an extra line to the line segments array which represents the last bottom row on the image
    last_bottom_row = np.flip(np.column_stack(((np.ones((img.shape[1],))*img.shape[0]), np.arange(img.shape[1]))).astype(int), axis=0)
    line_segments.append(last_bottom_row)

    line_images = []



    
    line_count = len(line_segments)
    fig, ax = plt.subplots(figsize=(10,10), nrows=line_count-1)
    output = []


    for line_index in range(line_count-1):
        line_image = extract_line_from_image(img, line_segments[line_index], line_segments[line_index+1])
        line_images.append(line_image)
        #print(line_image)
        #cv2.imwrite('/Users/vatsalya/Desktop/demo.jpeg',line_image)
        

        # im=Image.fromarray(line_image)
        # im=im.convert("L")
        # im.save("demo.jpeg")
        # print("#### Image Saved #######")
        imageio.imwrite('demo.jpeg',line_image)



        image = Image.open("demo.jpeg").convert("RGB")
        
        #print("Started Processing")
        #image = line_image
        pixel_values = processor(images=image, return_tensors="pt").pixel_values

        generated_ids = model.generate(pixel_values)
        generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
        print(generated_text)
        output.append(generated_text)
        #ax[line_index].imshow(line_image, cmap="gray")
    result=""
    for o in output:
        result=result+o
        result=result+" "
    return result



nlp = spacy.load("en_core_web_md")


def listToString(s):
   
    # initialize an empty string
    str1 = " "
   
    # return string 
    return (str1.join(s))

def rm_stop(my_doc):
    # Create list of word tokens
    token_list = []
    for token in my_doc:
        token_list.append(token.text)

    

    # Create list of word tokens after removing stopwords
    filtered_sentence =[] 

    for word in token_list:
        lexeme = nlp.vocab[word]
        if lexeme.is_stop == False:
            filtered_sentence.append(word)
    
    return filtered_sentence

def text_processing(sentence):

    sentence = [token.lemma_.lower()
                for token in nlp(sentence) 
                if token.is_alpha and not token.is_stop]
    
    return sentence

def jaccard_sim(sent1,sent2):
    # Text Processing
    sentence1 = text_processing(sent1)
    sentence2 = text_processing(sent2)
    
    # Jaccard similarity
    return textdistance.jaccard.normalized_similarity(sentence1, sentence2)

def sim(Ideal_Answer,Submitted_Answer):
# SBERT EMBEDDINGS
    text1=Ideal_Answer.replace("\"","").replace("\'","")
    text2=Submitted_Answer.replace("\"","").replace("\'","")
    output=[]
    model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')

    #Compute embedding for both lists
    embedding_1= model.encode(text1, convert_to_tensor=True)
    embedding_2 = model.encode(text2, convert_to_tensor=True)

    score=util.pytorch_cos_sim(embedding_1, embedding_2)
    output.append("SBERT:"+str(int(float(str(score).split("[")[2].split("]")[0])*10.0))+",")
    sbert=int(float(str(score).split("[")[2].split("]")[0])*10.0)
    #Jaccard
    output.append("Jaccard:"+str(int(jaccard_sim(text1,text2)*10.0))+",")

    #spacy average word2vec
    nlp = spacy.load("en_core_web_md")  # make sure to use larger package!
    doc1 =  listToString(rm_stop(nlp(text1)))
    doc2 =  listToString(rm_stop(nlp(text2)))

    # Similarity of two documents
    w2v=int(nlp(doc1).similarity(nlp(doc2))*10.0)
    final_score=int(0.8*sbert+0.2*w2v)
    output.append("Word2Vec:"+str(int(nlp(doc1).similarity(nlp(doc2))*10.0))+",final_score:"+str(final_score))
    out_string=listToString(output)
    #return out_string
    return str(out_string),final_score



def return_image_embedding(model,img_path):
    img = image.load_img(img_path, target_size=(224, 224))
    x = image.img_to_array(img)
    x = np.expand_dims(x, axis=0)
    x = preprocess_input(x)
    preds = model.predict(x)
    curr_df = pd.DataFrame(preds[0]).T
    return curr_df



def draw_boxes(image, bounds, color='yellow', width=2):
    draw = ImageDraw.Draw(image)
    for bound in bounds:
        p0, p1, p2, p3 = bound[0]
        draw.line([*p0, *p1, *p2, *p3, *p0], fill=color, width=width)
    return image

def inference(img, lang):
    reader = easyocr.Reader(lang)
    bounds = reader.readtext(img.name)
    im = PIL.Image.open(img.name)
    draw_boxes(im, bounds)
    im.save('result.jpg')
    return ['result.jpg', pd.DataFrame(bounds).iloc[: , 1:]]

def compute_tfidf_embeddings(documents1, documents2):
    # Combine both lists of words into a single list
    combined_documents = documents1 + documents2

    # Initialize the TF-IDF vectorizer
    vectorizer = TfidfVectorizer()

    # Fit the vectorizer on the combined documents
    vectorizer.fit(combined_documents)

    # Transform the documents to TF-IDF embeddings
    embeddings1 = vectorizer.transform(documents1)
    embeddings2 = vectorizer.transform(documents2)

    return embeddings1, embeddings2


def extract_eval(image1,image2,image3,image4):
    print(image1)
    ideal_text=extract(image1)
    print("Extracting Ideal Text \n")
    print(ideal_text)
    submitted_text=extract(image3)
    print("Extracting Submitted Text \n")
    print(submitted_text)
    a,b=sim(ideal_text,submitted_text)
    print(a)
    text_sim_score=b
    model = ResNet50(include_top=False, weights='imagenet', pooling='avg')
    diagram_1_embed=return_image_embedding(model,image2)
    diagram_2_embed=return_image_embedding(model,image4)
    diagram_embed_sim_score=util.pytorch_cos_sim(embedding_1, embedding_2)
    print("Diagram Embedding Similarity Score \n")
    print(diagram_embed_sim_score)
    
    

iface = gr.Interface(fn=extract_eval, 
                     #inputs=[gr.Image(label='Ideal Answer'),gr.Image(label='Ideal Answer Diagram'),gr.Image(label='Submitted Answer'),gr.Image(label='Submitted Answer Diagram')], 
                     #inputs=gr.inputs.File(file_count="directory"),
                     inputs=["image","image","image","image"],
                     outputs=gr.outputs.Textbox(),)

iface.launch(enable_queue=True)

# def preview(files, sd: gr.SelectData):
#     return files[sd.index].name

# with gr.Blocks() as demo:
#     with gr.Row():
#         with gr.Column():
#             f = gr.File(file_types=["image"], file_count="multiple")
#             i = gr.Image()
#             btn = gr.Button()
#         with gr.Column():
#             o = gr.Image()
    
#     f.select(preview, f, i)
#     btn.click(lambda x:x, i, o)
    
# demo.launch() abc