i found that there are some images/tensors in my dataset that result in torch.Size([4, 224, 224]) instead of torch.Size([3, 224, 224]), resulting in error RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 0. Got 3 and 4 in dimension 1. In the comment of pytorch it says that they swap color axis ©. now I want to of course change it from 4 to 3. what is the keyword for this? i use search machine with python image color axis but it spits out unrelated. is it related to rgb? so the image with size 4 in dimension 1 is not rgb?
class ToTensor(object):
"""Convert ndarrays in sample to Tensors."""
def __call__(self, sample):
image = sample['image']
# swap color axis because
# numpy image: H x W x C
# torch image: C X H X W
image2 = image.transpose((2, 0, 1)).copy()
tfm_image = torch.from_numpy(image2)
if image2.shape[0] != 3:
print("shape image = ", image.shape, " shape image2 = ", image2.shape, " tfm_image shape = ", tfm_image.shape)
return {'image': tfm_image,
'species_id': sample['species_id']}