Yolov3: visualize train_batch

[mehrdadazizi72]

Describe the bug
when I train on my custom dataset (tt100k), the first batch visualize by program but the bbox is not correct. I test bbox on my self and everything is right.

To Reproduce
the labels are :
9 0.1162109375 0.33837890625 0.07568359375 0.0986328125
90 0.10904931640625 0.2451171875 0.09087451171874998 0.0927734375

Screenshots

Desktop (please complete the following information):

• OS: ubuntu
• Version 18.04

Additional context
pytorch = 1.1.0
python = 3.7

[glenn-jocher]

@mehrdadazizi72 your labels are incorrect. The whole purpose of train_batch0.jpg is to see if they line up, and if they don't you've supplied incorrect labels. The bug is in your labels, not in our repo.

You can verify this by training on COCO:

[mehrdadazizi72]

@glenn-jocher thanks for your reply.
I visualize labels by my self that's show in the second picture and everything right.
also, write the labels if you want to check with a simple code. I think it's not my mistake but if you sure about that I will check it again. I will appreciate that you check labels in a simple code.
thanks for reply again.

[glenn-jocher]

@mehrdadazizi72 your labels are incorrect. Follow this tutorial:
https://github.com/ultralytics/yolov3/wiki/Train-Custom-Data