Hi,
Im looking at the colorjitter code and you apply the transformations like this:
for transform in transforms:
img_transformed = np.zeros(img.shape, dtype=np.float32)
for slice in range(img.shape[0]):
img_transformed[slice,:,:] = transform(img[slice,:,:].astype(np.float32))
return img_transformed
But this means you apply a transformation on img and keep zeroing out the img_transform after every transform applied. so basically the output is only the last transformation in the list.
Shouldnt it be something like:
img_transformed = deepcopy(img)
for transform in transforms:
for slice in range(img.shape[0]):
img_transformed[slice,:,:] = transform(img_transformed [slice,:,:].astype(np.float32))
return img_transformed
Hi, Im looking at the colorjitter code and you apply the transformations like this:
But this means you apply a transformation on img and keep zeroing out the img_transform after every transform applied. so basically the output is only the last transformation in the list. Shouldnt it be something like:
Thanks