openvinotoolkit / training_extensions

Train, Evaluate, Optimize, Deploy Computer Vision Models via OpenVINO™
https://openvinotoolkit.github.io/training_extensions/
Apache License 2.0
1.14k stars 443 forks source link

Feature Request: Add support for EfficientAD from Anomalib #3971

Open j99ca opened 2 months ago

j99ca commented 2 months ago

Currently there are only 2 supported anomaly models in OTX: PADIM and STFPM.

EfficientAD is a newer model that that Anomalib supports that has very high accuracy according to papers with code as well as my own internal testing for industrial use-cases. This would be a great model to include in OTX's anomaly detection flow, or perhaps tighter integration between OTX and Anomalib in general would be ideal.

harimkang commented 2 months ago

I personally would like to see all of Anomalib's models evolve to be commonly used in OTX, just like Timm or hugging-face models are supported by wrapper classes. But as far as I know, this needs some cleanup at the moment. But I think it's achievable. @ashwinvaidya17

ashwinvaidya17 commented 3 weeks ago

Yeah. That's the plan. We are going to slowly all support for all the algorithms available in Anomalib