Vadim

Hmm, when I remove all albumentations except HFlip, I still get such Error, but without any albumentations everything is OK.

My dataset has some bboxes with height or width 0. Therefore I modified them so they have W>1, H>1. After this training doesn't fail with albumentation: `HorizontalFlip` (UPDATE: no, it...

Currently fixed it by changing [this](https://github.com/albu/albumentations/blob/master/albumentations/core/composition.py#L156) line from `p.preprocess(data)` to ``` try: p.preprocess(data) except: print('data:', data) ``` This is костыль, and it doesn't augment some amount of images, but it...

Спасибо, что ответил по-китайски. Просто добавь volatile=True в строке https://github.com/yxlijun/Pyramidbox.pytorch/blob/master/tools/wider_test.py#L54

@joe-siyuan-qiao Thanks for the reply! Have you tried different epsilon values other than 1e-5 in `weight =/ (std + eps)`? E.g. will it degrade the quality if epsilon is bigger...

Thank you @richzhang for a quick response! 1. Suppose we use an **ensemble** of very deep CNNs as a teacher and do **knowledge distillation** to a small student CNN e.g....