Video models explained by finding minimal sets of critical frames
Sufficient Frame Subsets for Video Understanding
Computer Vision and Pattern Recognition
Summary
Video understanding models often rely on specific moments in a video to make decisions, but it's hard to tell exactly which parts matter most. The authors developed a method to find small groups of key frames that alone can explain the model’s prediction without needing the rest of the video. By testing on popular video datasets and different model types, they showed their method picks very compact and meaningful frames. This helps understand how models learn, spots where they go wrong, and suggests ways to improve future training.
What this means in practice
- •For machine learning engineers: Identify which frames a video model depends on to improve its reliability and fix errors in video classification systems.
- •For video analytics developers: Streamline video content processing by focusing on critical frames that maintain model accuracy, reducing computational cost.
Authors
Lahiru Wijayasingha
Abstract
Understanding when during a video a model makes its critical observations is essential for interpreting video-understanding systems. However, existing attribution methods often spread evidence diffusely across near-duplicate frames. We introduce Sufficient Frame Subsets: compact sets of frames in which all other frames are masked, yet each subset independently preserves the model's prediction. We evaluate our method on two complementary benchmarks, UCF101 and Something-Something v2, using both CNN- and transformer-based architectures. Our results show that the proposed algorithm identifies highly compact and effective frame subsets for explaining model decisions. We also find that models capable of relying on a greater number of sufficient frame subsets tend to perform better, possibly because they are more robust, consistent with prior research. Finally, the extracted frame subsets help us analyze model failure cases, reveal which aspects of a video the model has learned to emphasize, and identify opportunities for improving future training procedures.