anomalib icon indicating copy to clipboard operation
anomalib copied to clipboard

Published 20 hours ago • verified publisher icon openvinotoolkit

Why is the loss function of fastflow negative, What are the meanings of pixel_F1Score and pixel_AUROC？Thanks

Open leibaoer opened this issue 2 years ago • 1 comments

image

leibaoer avatar Sep 14 '22 02:09 leibaoer

hello

leibaoer avatar Sep 20 '22 01:09 leibaoer

The loss is likelihood as general normalizing flow models. It will be helpful to search concepts of normalizing flow. In addition, fastflow predicts whether the pixel is anomaly. Pixel AUROC means the localization(segmentation) perfomance.

cytotoxicity8 avatar Sep 23 '22 14:09 cytotoxicity8

@leibaoer sorry for the late response. As @cytotoxicity8 mentioned, fastflow computes the loss by maximizing the negative log jacobian. You can see it defined here https://github.com/openvinotoolkit/anomalib/blob/main/anomalib/models/fastflow/loss.py#L27. Section 3.3 of the fastflow paper (https://arxiv.org/pdf/2111.07677.pdf) might make it more clear.

ashwinvaidya17 avatar Sep 23 '22 14:09 ashwinvaidya17