search

nuclearboy95 / Anomaly-Detection-PatchSVDD-PyTorch

248 stars 56 forks source link

Over-sensitive #25

Open metalluca opened 1 year ago

metalluca commented 1 year ago

Hello everyone,

Thank you very much for the interesting work!

I trained the algorithm on a custom dataset containing mostly sensory and texture-like data. The results contain many falsely identified anomalies (see second images) in normal data. Do you have any suggestion on preventing this over-sensitive behavior? Thanks in advance!

image image

nuclearboy95 commented 1 year ago

Hi, it seems to be a normalization problem (checkout issue #9 and #22 ).

The visualization code performs min-max normalization using the min and max values in the heatmap.

For a clear visualization, normalize the maps using the min and max values in the heatmaps of the entire dataset.

metalluca commented 1 year ago

Thank you for the quick response!