ValueError: Height and Width of image, mask or masks should be equal.disable shapes check by setting is_check_shapes=False of Compose class

Question

I would like to mention that the shape of image before transformation is:

	Shape
Original image	(360, 640, 3)
Original mask	(360, 640)
Transformed image	(320, 320, 3)
Transformed mask	(320, 320)

and img_trans is defined as

img_trans = A.Compose([
    A.Resize(a_config.INPUT_IMAGE_HEIGHT, a_config.INPUT_IMAGE_WIDTH)],
    is_check_shapes=False
)

The code snippet is:

def image_mask_transformation(image, mask, img_trans, aug_trans=None, normalize=True,
                              var_min=10, var_max=400):

    
    
    print("Original image shape:", image.shape)
    transformed = img_trans(image=image, mask=mask)
    image = transformed["image"]
    mask = transformed["mask"]
    print("transformed image shape:", image.shape)
    # image, mask still uint 8 in 0, 255 range

But I am getting an error:

ValueError: Caught ValueError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/worker.py", line 308, in _worker_loop
    data = fetcher.fetch(index)
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/fetch.py", line 51, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/fetch.py", line 51, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/content/drive/MyDrive/Colab Notebooks/CRACK500/tool/dataset.py", line 131, in __getitem__
    image_store, mask_store = image_mask_transformation(image, mask, self.img_trans, aug_trans=self.aug_trans,
  File "/content/drive/MyDrive/Colab Notebooks/CRACK500/tool/dataset.py", line 40, in image_mask_transformation
    transformed = img_trans(image=image, mask=mask)
  File "/usr/local/lib/python3.10/dist-packages/albumentations/core/composition.py", line 195, in __call__
    self._check_args(**data)
  File "/usr/local/lib/python3.10/dist-packages/albumentations/core/composition.py", line 286, in _check_args
    raise ValueError(
ValueError: Height and Width of image, mask or masks should be equal. You can disable shapes check by setting a parameter is_check_shapes=False of Compose class (do it only if you are sure about your data consistency)..

I am not sure whether I am getting the error because initially a 3rd dimension is there for image and not for the mask. Hence I have tried to add the is_check_shapes=False while defining img_trans but I am still getting the error.

Can you suggest whether and where I should add is_check_shapes=False or the error can be solved in any other way?

Answer 1

This is old now, but I had a problem where I was changing the dimensions of my image/mask before I performed the transforms. In my case, I was converting my images to channel first before transforming in my pytorch Dataset.

I had this:

def __getitem__(self, idx):
  ...
  if self.channel_first:
        image = np.transpose(image, (2, 0, 1)).astype(np.float32)
        mask = np.transpose(mask, (2, 0, 1)).astype(np.float32)

    if self.transform:
        transformed = self.transform(image=image, mask=mask)
        image = transformed["image"]
        mask = transformed["mask"]

when I should have had this:

def __getitem__(self, idx):
   ...

    if self.transform:
        transformed = self.transform(image=image, mask=mask)
        image = transformed["image"]
        mask = transformed["mask"]

    if self.channel_first:
        image = np.transpose(image, (2, 0, 1)).astype(np.float32)
        mask = np.transpose(mask, (2, 0, 1)).astype(np.float32)