Extending Information Bottleneck Attribution to Video Sequences

TL;DR

VIBA is a new explainability method for video classification that adapts Information Bottleneck Attribution to highlight manipulated regions and motion inconsistencies in deepfake detection, providing consistent and human-aligned explanations.

Contribution

This work extends IBA to video sequences, enabling explainability in temporal models for video analysis, especially for deepfake detection.

Findings

VIBA produces temporally and spatially consistent explanations.

VIBA's relevance maps align closely with human annotations.

Effective in highlighting manipulated regions and motion inconsistencies.

Abstract

We introduce VIBA, a novel approach for explainable video classification by adapting Information Bottlenecks for Attribution (IBA) to video sequences. While most traditional explainability methods are designed for image models, our IBA framework addresses the need for explainability in temporal models used for video analysis. To demonstrate its effectiveness, we apply VIBA to video deepfake detection, testing it on two architectures: the Xception model for spatial features and a VGG11-based model for capturing motion dynamics through optical flow. Using a custom dataset that reflects recent deepfake generation techniques, we adapt IBA to create relevance and optical flow maps, visually highlighting manipulated regions and motion inconsistencies. Our results show that VIBA generates temporally and spatially consistent explanations, which align closely with human annotations, thus providing interpretability for video classification and particularly for deepfake detection.