XAI saliency/feature maps removed in v3.0

[yvolokitin]

We want to let the community know about a change to Explainable AI (XAI) functionality in the current release.

Description:
What changed: Explainable AI visualizations — saliency maps and feature maps — are not available in Geti v3.0.

Why: The existing saliency algorithms do not generalize reliably across the diverse set of model architectures now supported in v3.0, and in some cases can produce misleading output. Rather than ship visualizations that may not be trustworthy, we've removed the feature for this release.

Impact: No saliency or feature map visualization is available in v3.0.

Action required:
None. There is no equivalent available in v3.0. A reworked version is under consideration for a future release.

Feedback
If XAI/saliency is important to your workflow, please share your interest and use cases in the tracking thread
please share your interest and use cases in the tracking thread (#6588)