Explaining Motion Relevance for Activity Recognition in Video Deep Learning Models

TL;DR

This paper introduces a method to adapt 2D explanation techniques for 3D activity recognition models, enabling clearer understanding of motion's role in model decisions, which was previously obscured.

Contribution

The authors propose a selective relevance method that enhances 2D explanation techniques to specifically highlight motion information in 3D activity recognition models.

Findings

Improves explanation selectivity for motion in 3D models

Reveals and quantifies the model's spatial bias

Simplifies explanations for human understanding

Abstract

A small subset of explainability techniques developed initially for image recognition models has recently been applied for interpretability of 3D Convolutional Neural Network models in activity recognition tasks. Much like the models themselves, the techniques require little or no modification to be compatible with 3D inputs. However, these explanation techniques regard spatial and temporal information jointly. Therefore, using such explanation techniques, a user cannot explicitly distinguish the role of motion in a 3D model's decision. In fact, it has been shown that these models do not appropriately factor motion information into their decision. We propose a selective relevance method for adapting the 2D explanation techniques to provide motion-specific explanations, better aligning them with the human understanding of motion as conceptually separate from static spatial features. We demonstrate the utility of our method in conjunction with several widely-used 2D explanation methods, and show that it improves explanation selectivity for motion. Our results show that the selective relevance method can not only provide insight on the role played by motion in the model's decision -- in effect, revealing and quantifying the model's spatial bias -- but the method also simplifies the resulting explanations for human consumption.