You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It would be great if Model API models had a method similar to the postprocess, that would generate a saliency map for models in any domain. Currently, we need to keep that logic on the GETi side. It exists in Mediapipe and serves the WEB UI and now I need to implement it in Python for GETi SDK.
Including, for segmentation models, it would be nice to get activations before softmax, resulting in softer maps.
And preparing saliency maps for anomaly models as well.
The text was updated successfully, but these errors were encountered:
@igor-davidyuk there are no plans to add XAI algorithms in MAPI. Only models with already integrated XAI modules are supported. Instead, we have a standalone project that can embed XAI branches into already generated IR models https://github.com/openvinotoolkit/openvino_xai
It would be great if Model API models had a method similar to the
postprocess
, that would generate a saliency map for models in any domain. Currently, we need to keep that logic on the GETi side. It exists in Mediapipe and serves the WEB UI and now I need to implement it in Python for GETi SDK.Including, for segmentation models, it would be nice to get activations before softmax, resulting in softer maps.
And preparing saliency maps for anomaly models as well.
The text was updated successfully, but these errors were encountered: