GT for train

[rzamarefat]

Hi, thank you for the implementation. I want to train this repo on some other languages but I am stuck in a problem. In the repo you have explained how one can provide the model the gts? In "ICDAR2015.txt" file you have just the paths to the images but no bounding boxes ground truths. Do I have to make another txt file for storing bounding boxes for each image? and If yes in what format should it be?
for instance:

path/to/image/ (200, 300,400,500,300,400, 600,700) which demonstrates the four corners of just one image.