DragonZzzz

label and num_classes of the human parseing

1

Hi, author. I want to generate training data using a custom dataset. I find that LIP has 20 classes and ATR has 18 classes. `LIP: 'Background', 'Hat', 'Hair', 'Glove', 'Sunglasses',...

How to get the evaluation results of Facewarehouse and MICC Florence?

2

How to get the evaluation results of Facewarehouse and MICC Florence? And are there differences compared with other methods in evaluation process? Could you provide an evaluation code? Thank you...

Random seed Problem

3

作者您好，从代码中我发现您使用设置的随机种子time.time()，这会导致每次训练时初始化情况不一样。在训练时，我发现在不同的时间开始训练可能会影响网络的收敛情况，从而直接影响网络得到的效果，所以并没有得到作者在论文中提出的效果。请问作者在训练时会遇到这个问题吗？

The problem of training

3

When I train BSDS500 dataset, the images in tmp are black, and the loss is almost zero. The training has crashed.

How to use the utils for our depth images?

2

@s-gupta Sorry to botter you, but I hava a question for consultation. I hava compiled the imagestack library. If I want to use the rgbdutils to change depth images to...