How to convert polygon coordinates to rectangle(yolo format) for an image labelling?-CodePudding

I am trying to read water meter reading through OCR, however, my first step is to find ROI. I found a dataset from Kaggle with the labeled data for the ROI. But they are not in rectangles, rather in a polygon shape, some with 5 points, and some with 8 depending on the image. How do I convert this to yolo format? For example:

file name | value | coordinates

id_53_value_595_825.jpg 595.825 {'type': 'polygon', 'data': [{'x': 0.30788, 'y': 0.30207}, {'x': 0.30676, 'y': 0.32731}, {'x': 0.53501, 'y': 0.33068}, {'x': 0.53445, 'y': 0.33699}, {'x': 0.56529, 'y': 0.33741}, {'x': 0.56697, 'y': 0.29786}, {'x': 0.53501, 'y': 0.29786}, {'x': 0.53445, 'y': 0.30417}]}

id_553_value_65_475.jpg 65.475 {'type': 'polygon', 'data': [{'x': 0.26133, 'y': 0.24071}, {'x': 0.31405, 'y': 0.23473}, {'x': 0.31741, 'y': 0.26688}, {'x': 0.30676, 'y': 0.26763}, {'x': 0.33985, 'y': 0.60851}, {'x': 0.29386, 'y': 0.61449}]}

id_407_value_21_86.jpg 21.86 {'type': 'polygon', 'data': [{'x': 0.27545, 'y': 0.19134}, {'x': 0.37483, 'y': 0.18282}, {'x': 0.38935, 'y': 0.76071}, {'x': 0.28185, 'y': 0.76613}]}

I understood that for yolo format, I need to get xmin, ymin, xmax, ymax so that i can calculate the width and height but i have trouble with parsing the data. Could anyone help?

Thank you.

Edit: Finally, it worked out. Incase anyone is struggling with converting csv file to yolo format from https://www.kaggle.com/datasets/tapakah68/yandextoloka-water-meters-dataset, here is my code snipet to just create text files for each image.

import csv
import pandas as pd
import json
import ast
def converttoyolo(csv_file):
  df = pd.read_csv(csv_file)
  l_csv = len(df)
  for i in range(l_csv):
    df_row = df.iloc[i]  #get each row
    
    df_ = df_row['photo_name']  #image column 
    df__ = df_.split('.')  #to get name for text file

    df_new = df_row['location']  #start of gettinf coordinates access
    df_dict = ast.literal_eval(df_new)  #str to dict
    df__dict = json.dumps(df_dict, indent = 4)
    df_dict__ = json.loads(df__dict)
    
    convertedDict = df_dict__
    length = len(convertedDict['data'])
    x = []
    y = []
    for j in range(length):  #put each x and y for each row in seperate array
      x.append(convertedDict['data'][j]['x'])
      y.append(convertedDict['data'][j]['y'])

    max_x = max(x)
    max_y = max(y)  #yolo conversion, check answer below
    min_x = min(x)
    min_y = min(y)

    width = max_x - min_x
    height = max_y - min_y 
    center_x = min_x   (width/2)
    center_y = min_y   (height/2)

    def filename(file):   #put in text files
      with open(file ".txt", "w") as file:
        
         file.write(str(width) ',' str(height) ','  ...
         str(center_y) ',' str(center_y))
    
    filename('/content/drive/MyDrive/yolo/custom_data/jpeg/' df__[0]) 
    
converttoyolo(csv_file)

CodePudding user response：

You need to create a contour (a list of points) for each shape. Once you have that, then call cv::boundingRect() to turn each contour into a a single bounding rectangle. Once you have the rectangle, then you you can figure out X, Y, W, and H. But since YOLO format is CX and CY -- not X and Y -- then you need to do:

CX = X   W/2.0
CY = Y   H/2.0

Lastly, you must normalize all 4 values. The YOLO format is space delimited, and the first value is the integer class ID. So if "dog" is your 2nd class (thus id #1 since it is zero-based), then you'd output:

1 0.234 0.456 0.123 0.111

...where the 4 coordinates are:

CX / image width
CY / image height
W / image width
H / image height

If you want more examples of the math, see the Darknet/YOLO FAQ: https://www.ccoderun.ca/programming/darknet_faq/#darknet_annotations