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

Inference codes and trained weights ? #1

Open opentld opened 2 years ago

opentld commented 2 years ago

RT :)

Atotem commented 1 year ago

Hi @opentld ,

Apparently, to perform an inference you technically need, from what I understand, an image with a seen anomaly, an image with a pseudo-anomaly and a reference image. Which for me makes no sense at all to ask for all that kind of data just to make an inference. Maybe there's a reason for this which could be very helpful to have it explained from the original writers, which I doubt.

Screenshot 2022-11-17 114512 (Choubo Ding et al, page 5 line 7)