unable to train Error processing 100_icdar13.png .What split_label.py exactly dose and what is the correct input format for image and label inside dataset directory

[AjitAntony]

hi ,

im trying to train with custom data and after running utils/prepare/split_label.py i get below unable to train Error processing 100_icdar13.png .

Folder structure dataset ---custom_image---image--- 100_icdar13.png ---custom_image---label---gt_100_icdar13.txt --image (empty folder) --label (empty folder)

When i check the code the code is going to exception apart after the 2nd line .

for line in lines: splitted_line = line.strip().lower().split(',') x1, y1, x2, y2, x3, y3, x4, y4 = map(float, splitted_line[:8])

i went head and printed the splitted_line of the 1st line in the gt_100_icdar13.txt file splitted_line ['201', '162', '207', '229'] ['201', '162', '207', '229']

this line is failing x1, y1, x2, y2, x3, y3, x4, y4 = map(float, splitted_line[:8]) due to not enough values to unpack (expected 8, got 4)

SO IS IT MANDATORY FOR THE LABEL.TXT FILE TO HAVE MORE THAN 8 VALUES ? AND ARE THESE VALUES REPRESENTING THE CO-ORDINATES OF THE TEXT ? WHY SHOULD WE GIVE THE CO-ORDINATES OF THE TEXT ALONG WITH IMAGE ? because split_label.py also outputs and image with the lable

[sevany]

hey, do you solve this problem yet?