ENH: Partition Explainer for Video Models

[hex-plex]

Problem Description

While DeepExplainer can be better for the problem, with newer black box models demonstrating higher accuracy on Video Classification, Question Answering and Video Captioning, it would be interesting to learn the intrinsic behaviour of these models for Hallucination and susceptibility. I wanted to start out with the basic video classification models but the Image Masker are fixed for 2D inputs and require multiple alterations to adapt for a video.

Alternative Solutions

• Extend the Partition tree to support video
• Extend masker and Blur functionalities to support videos
• Fix internal Plot to support video plotting
• Optimize for both script/notebook execution

Additional Context

These are 3d plots for the implementation

for the following video

Feature request checklist

• I have checked the issue tracker for duplicate issues.
• I'd be interested in making a PR to implement this feature