search

Choubo / DRA

Official PyTorch implementation of the paper “Catching Both Gray and Black Swans: Open-set Supervised Anomaly Detection”, open-set anomaly detection, few-shot anomaly detection.
GNU Affero General Public License v3.0
72 stars 11 forks source link

locate the defect areas #2

Open jiyidefeng1990 opened 2 years ago

jiyidefeng1990 commented 2 years ago

really appreciate your great work. One question, could you tell me how to locate the defect areas of the anomaly sample, for example, give a rectangle to the defect areas like most object detection methods did?

Atotem commented 1 year ago

@jiyidefeng1990 Hi! Did you manage to render some output? Or run a test script? This could be very useful. Best regards!

Choubo commented 1 year ago

Hi, @jiyidefeng1990

Please refer to DevNet, using gradients to detect defect regions is a common anomaly localization method.

Cheers, Choubo