get_dataframe.py

import cv2
import pandas as pd
from paddleocr import PaddleOCR


def get_dataframe(match_path):
    
    dataframe = {
        "Timestamp":[],
        "Team1":[],
        "Team2":[],
        "Score1": [],
        "Score2":[]
    }
    
    cap = cv2.VideoCapture(match_path)
    i = 0 
    frame_skip = 200
    frame_count = 0

    ocr = PaddleOCR(use_angle_cls=True, lang='en')
    while cap.isOpened():
        timestamp = cap.get(cv2.CAP_PROP_POS_MSEC)
        timestamp = timestamp//1000
        ret, frame = cap.read()
        if not ret:
            break
        if i > frame_skip -1:
            frame_count += 1
            cropped_frame = frame[40:65,110:313]
            
           
            try:
                result = ocr.ocr(cropped_frame, cls=True, det=False)
            except Exception as e:
                print("Error during OCR:", e)
                continue
            for line in result:
                for text, conf in line:
                    if conf > 0.90:
                        main_text = text.strip()
                        break
                else:
                    continue
                if main_text[:3] != ' ':
                    team1 = main_text[:3]
                    dataframe['Team1'].append(team1)
                if main_text[4:5] != ' ':
                    score2 = main_text[4:5]
                    dataframe['Score2'].append(score2)
                if main_text[5:] != ' ':
                    team2 = main_text[5:]
                    dataframe['Team2'].append(team2)
                if main_text[3] != ' ':
                    score1 = main_text[3:4]
                    dataframe['Score1'].append(score1)
                dataframe['Timestamp'].append(timestamp)                        
            i =0
            print('------GENERATING DATA FRAME------')
            continue
        i +=1
        
    df_final = pd.DataFrame.from_dict(dataframe)
   
    cap.release()
    cv2.destroyAllWindows()
    return  df_final.to_csv('file1.csv',index=False)