Play Fair: Frame Attributions in Video Models

Abstract

In this paper, we introduce an attribution method for explaining action recognition models. Such models fuse information from multiple frames within a video, through score aggregation or relational reasoning. We break down a model's class score into the sum of contributions from each frame, fairly. Our method adapts an axiomatic solution to fair reward distribution in cooperative games, known as the Shapley value, for elements in a variable-length sequence, which we call the Element Shapley Value (ESV). Critically, we propose a tractable approximation of ESV that scales linearly with the number of frames in the sequence. We employ ESV to explain two action recognition models (TRN and TSN) on the fine-grained dataset Something-Something. We offer detailed analysis of supporting/distracting frames, and the relationships of ESVs to the frame's position, class prediction, and sequence length. We compare ESV to naive baselines and two commonly used feature attribution methods: Grad-CAM and Integrated-Gradients.

Code available at: https://github.com/willprice/play-fair/ and supporting website at: https://play-fair.willprice.dev/

References in corpus (9)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Grad-CAM++: Improved Visual Explanations for Deep Convolutional Networks
• Axiomatic Attribution for Deep Networks
• The Kinetics Human Action Video Dataset
• Learning Important Features Through Propagating Activation Differences
• YouTube-8M: A Large-Scale Video Classification Benchmark
• Sanity Checks for Saliency Maps
• Spatiotemporal Residual Networks for Video Action Recognition
• RISE: Randomized Input Sampling for Explanation of Black-box Models