choose the right setting for custom anomaly detection with unsupervised nature.

[hrnbot]

I want to find anomalies unsupervised in the below dataset, I have tried all the algorithms of anomalib but padim only worked but it was also not able to detect the crack.
NG品㈫場所.xlsx

NG品キズ場所.xlsx
NG品黒点場所.xlsx
NG品㈬場所.xlsx
NG品㈪場所.xlsx

[ashwinvaidya17]

Can you provide more details (you can take inspiration from our issue template? Did you provide segmentation mask annotations for your dataset? This helps compute pixel-level thresholds that increase the accuracy of predictions.

Also, can you attach the images directly rather than embedding them in an xlsx file.

[hrnbot]

I have only few (5) NG images. others are good images

[hrnbot]

I have 100 good images, i am using custom dataset as mentioned above which are anomalies.