|
|
|
if self._representation == 'iframe': |
|
img = color_aug(img) |
|
|
|
# BGR to RGB. (PyTorch uses RGB according to doc.) |
|
img = img[..., ::-1] |
|
|
|
frames.append(img) |
When creating val_dataloader, is_train is set as False. But color jittering is still enabled. This may end up with inconsistent test results.