- Use the train_PNI method (in patchcore.py) to train the MLP and record histograms any number of epochs
- Use the PNI_predict (in patchcore.py) method for (obviously) predicting
- Use the refinement_model.py which is the exact refinement architecture as the paper
- Dataloader used here is a custom dataloader (not MVTech)
- There is also a custom cutpaste transformation that also generates GTs
amroabuzer / pni-anomaly-detection Goto Github PK
View Code? Open in Web Editor NEWThis is an unofficial implementation of the paper PNI : Industrial Anomaly Detection.