Set transformations from the config file

[alexriedel1]

This PR allows to define image augmentations and transforms that are part of the albumentations module to be defined in the model config file.

This becomes possible now:

dataset:
  name: mvtec
  format: mvtec
  path: ./datasets/MVTec
  task: segmentation
  category: bottle
  train_batch_size: 32
  test_batch_size: 32
  num_workers: 8
  image_size: 256 # dimensions to which images are resized (mandatory)
  center_crop: 224 # dimensions to which images are center-cropped after resizing (optional)
  normalization: imagenet # data distribution to which the images will be normalized: [none, imagenet]
  transform_config:
    train:
      Resize:
        height: 256
        width: 256
        always_apply: true
      Normalize:
        mean: [0.485, 0.456, 0.406]
        std: [0.229, 0.224, 0.225]
    eval:
      Resize:
        height: 256
        width: 256
        always_apply: true
      Normalize:
        mean: [0.485, 0.456, 0.406]
        std: [0.229, 0.224, 0.225]

It not only helps readability but also allows for image augmentations to be used in HPO sweeps

[ashwinvaidya17]

Thanks! This is very useful

[djdameln]

Can you explain the changes to this file?

[alexriedel1]

If you perform HPO sweeps on colab, the GPU memory isn't freed after each run and accumulates over time. To solve this, the model needs to be deleted and torch cache has to be cleared:
https://stackoverflow.com/a/57569472/11951277

[samet-akcay]

@alexriedel1, not sure you noticed, but pre-commit checks failed due to the following error

src/anomalib/utils/hpo/runners.py:6:1: I001 [*] Import block is un-sorted or un-formatted
Found 1 error.

[alexriedel1]

@alexriedel1, not sure you noticed, but pre-commit checks failed due to the following error

src/anomalib/utils/hpo/runners.py:6:1: I001 [*] Import block is un-sorted or un-formatted
Found 1 error.

yess, working on it :)

[alexriedel1]

from the failing pre-commit checks:

[2023-04-04T09:58:21.403Z] ['error'] There was an error running the uploader: 
Unable to detect SHA and slug, please specify them manually.

Thats not on my side right?

[samet-akcay]

from the failing pre-commit checks:

[2023-04-04T09:58:21.403Z] ['error'] There was an error running the uploader: 
Unable to detect SHA and slug, please specify them manually.

Thats not on my side right?

No, it sometimes happens when the coverage is uploaded to codecov. The tests pass. We could merge this.

Thanks again! :)

[sudhanv09]

Hi can you give some more examples of how to do this. I have been searching for transform config but to no avail. I found your method but the yaml parser always fails to parse. Thanks

[alexriedel1]

Hi can you give some more examples of how to do this. I have been searching for transform config but to no avail. I found your method but the yaml parser always fails to parse. Thanks

Do you have installed the latest version? What exact errors are you getting?

[sudhanv09]

Ah my bad i was on the release version, now I have installed this dev version, it works. Thanks