Faithful analysis method explains video anomaly detector decisions
How to Explain your Unexplainable Video Anomaly Detector
Computer Vision and Pattern Recognition
Summary
Detecting unusual events in videos is often done by complex models that are hard to understand. The authors introduce a way to explain what parts of a video’s features a fixed anomaly detector relies on by making small, targeted changes to those features and seeing how the detector’s outputs change. Their method represents video features as connected parts like agents, actions, and outcomes, then isolates each part to see its impact. They find that changing certain elements, especially outcomes, strongly affects detector scores, proving their approach helps uncover what the detector truly uses to spot anomalies.
What this means in practice
- •For video surveillance teams: Diagnose and explain decisions of fixed video anomaly detectors to improve trust and troubleshooting during security monitoring.
- •For autonomous vehicle developers: Support understanding of anomaly detection modules by auditing feature dependencies without retraining existing detectors.
Tested on one dataset.
Authors
Giacomom D'Amicantonio, Egor Bondarev
Abstract
Weakly-supervised Video Anomaly Detection (VAD) models assign anomaly scores from high-dimensional features, and their predictions remain unexplained. Saliency maps show where a model looks but not what it relies on, and the descriptions produced by Vision-Language Models (VLMs) are generated by branches that the anomaly score never reads. To address this faithfulness gap, we propose \textbf{Explain The Unexplainable (ETU)}, a classifier-agnostic framework that audits frozen VAD models through counterfactual interventions in their feature space. ETU maps video features to a discrete dependency graph of agents, actions and outcomes ($A_r \to A_a \to A_o$), regenerates the features by graph-conditioned diffusion with selected nodes of the graph masked, and measures the change in the score of a frozen detector; one set of interventions is applied unchanged to every detector. The generator is trained only to reconstruct the input features, so it cannot learn the contrast it is asked to measure, and we validate the audit with experiments that test every alternative explanation of its results, from the cost of regeneration to random deletions of matched size. On UCF-Crime features obtained from a CLIP encoder, we show that substituting the scene semantics results in the largest drop for most detectors, and that removing the outcome chain of the graph ($Δ_{rel}$) changes the anomaly score of all six audited detectors far more than removing the same number of random nodes, even though the random deletion perturbs the features about as much. We show that this targeted behavior comes from the divergence objective of the graph embedding. Without it, the two interventions become indistinguishable, and with a divergence objective in the generator as well, the score change of the detector follows the displacement of the reconstruction rather than the semantics that were removed.