When test_split is applied to a custom dataset, test data contamination occurs

[jeongHwarr]

Describe the bug

• When test_split is applied to a custom dataset, test data contamination occurs. (test data(test data is also used for training).

To Reproduce
Steps to reproduce the behavior:

1. Use custom dataset
2. Set split_ratio to a specific value in config.yaml (without nomal_test_dir)
3. See error (test data contamination)

Expected behavior

• I hope that the train and test samples are the same because the same seed is used.
  I hope that the train and test results are the same because the same seed is used.
  https://github.com/openvinotoolkit/anomalib/blob/c5220c037f1580df5f7a7474ae488483f0f83eae/anomalib/data/folder.py#L164
• However, samples are different. So test data contamination occurs.
  
  

Screenshots

• If applicable, add screenshots to help explain your problem.

Hardware and Software Configuration

• OS: [ Windows 10 ]
• NVIDIA Driver Version [470.57.02]
• CUDA Version [11.3.1]

Additional context

• PatchCore