AdneneBoumessouer / MVTec-Anomaly-Detection

This project proposes an end-to-end framework for semi-supervised Anomaly Detection and Segmentation in images based on Deep Learning.
278 stars 59 forks source link

Dev2 #16

Closed AdneneBoumessouer closed 4 years ago

AdneneBoumessouer commented 4 years ago

This validation algorithm which determines the threshold when given an area (a number of pixels) as a parameter is now complete. This area represents the minimum defect area that a connected component in the thresholded anomaly map must have to be classified as a defective region. This assumes that the user defined area is a good value, which may not always be the case. TO DO NEXT: inverstigate a good systemmatic way to approximate this area.