shellysheynin / a-hierarchical-transformation-discriminating-generative-model-for-few-shot-anomaly-detection Goto Github PK

Is it possible you can upload the model weights somewhere?

I have been playing with this repo for a while now, and could train both CIFAR, and MNIST, with great evaluation performance.

The problem is in the Paris dataset (DataLoader, maybe?) - while it produces 5 scales in training (0-4), as for MNIST and CIFAR. I expect it to evaluate on these scales, and only these (0-4)

The problem is that while I am trying to evaluate Paris dataset - it looks for a 6th scale (6) and thus crashes (couldn't find the 6th pth file)

The fix might be very simple, but first I want to make sure that I am not missing something.

Does anyone have any insights about this issue?

Would you mind choosing a license for how you'd like your work to be used? Thanks

Hi,
Should we simply remove 'opt.num_images = 1' in Line 45 in defect_detection_evaluation.py?