ValueError: Requested crop size is larger than the image size in RandomResizedCrop

🐛 Bug

ValueError: Requested crop size is larger than the image size when including RandomResizedCrop in my transforms pipeline:

ValueError: Caught ValueError in DataLoader worker process 1.
Original Traceback (most recent call last):
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/_utils/worker.py", line 202, in _worker_loop
    data = fetcher.fetch(index)
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/workspace/ai_damages/parts/src/data_loading.py", line 64, in __getitem__
    transformed = self.transforms(image=img, mask=mask)
  File "/opt/conda/lib/python3.8/site-packages/albumentations/core/composition.py", line 182, in __call__
    data = t(force_apply=force_apply, **data)
  File "/opt/conda/lib/python3.8/site-packages/albumentations/core/transforms_interface.py", line 89, in __call__
    return self.apply_with_params(params, **kwargs)
  File "/opt/conda/lib/python3.8/site-packages/albumentations/core/transforms_interface.py", line 102, in apply_with_params
    res[key] = target_function(arg, **dict(params, **target_dependencies))
  File "/opt/conda/lib/python3.8/site-packages/albumentations/core/transforms_interface.py", line 227, in apply_to_mask
    return self.apply(img, **{k: cv2.INTER_NEAREST if k == "interpolation" else v for k, v in params.items()})
  File "/opt/conda/lib/python3.8/site-packages/albumentations/augmentations/transforms.py", line 857, in apply
    crop = F.random_crop(img, crop_height, crop_width, h_start, w_start)
  File "/opt/conda/lib/python3.8/site-packages/albumentations/augmentations/functional.py", line 331, in random_crop
    raise ValueError(
ValueError: Requested crop size (2131, 1636) is larger than the image size (1836, 3264)

To Reproduce

I can’t reproduce it deterministically, it just happens sometimes.

Environment

• Docker Image: pytorch/pytorch:1.8.1-cuda11.1-cudnn8-runtime
• Albumentations version: 0.5.2 (by pip)

Additional Context

Instantiated as: A.RandomResizedCrop(img_size, img_size, scale=(0.20, 1.))

[Dipet]

Now we check that image and mask shapes are equal. More info in #1310

[lucmos]

@claverru Thanks! I hit almost exactly the same problem 😃