mlsanityData to 14 Oct 2026

Look Inside Each Video: Rethinking How Video Anomaly Detection Is Evaluated

arXiv 2608.21854 ↗First posted 22 Aug 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Video anomaly detection (VAD) aims to localize anomalous events by identifying when they occur. A common evaluation pools frame-level anomaly scores across all test videos and measures whether anomalous frames receive higher scores than normal frames (Micro-AUROC). This pooling creates comparisons both within the same video and across different videos. However, the pooled evaluation protocol commonly used in current VAD benchmarks makes Micro-AUROC heavily dependent on cross-video comparisons, which do not directly test whether anomalous frames rank above normal frames within the same video. To quantify this imbalance, we decompose Micro-AUROC into within-video and cross-video comparisons. Across four widely used weakly supervised VAD benchmarks, within-video pairs account for only 0.071-0.388% of all anomalous-normal frame pairs. This imbalance can make Micro-AUROC a poor indicator of within-video ranking. Video-level classifiers that assign a single score to each video achieve 81.40-97.18% Micro-AUROC while obtaining 50% Within-AUROC. For existing weakly supervised VAD detectors, replacing all frame scores in each video with their mean sets Within-AUROC to 50%. Across 120 runs, this replacement retains a median 98.8% of the original Micro-AUROC margin above chance. Because the decomposition depends on the evaluation protocol rather than the training supervision, the same distinction applies whenever VAD frame scores are pooled across videos. High Micro-AUROC alone therefore does not establish temporal localization. Our analysis reveals a limitation of pooled VAD evaluation and motivates reporting within-video ranking alongside Micro-AUROC.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 77 of its 119 references that are in the map. First posted on arXiv 22 Aug 2026.

Builds on

77 of 119 references.

References · 77

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2017Quo Vadis, Action Recognition? A New Model and the Kinetics Datasetcited by 3,270
  3. 2018Real-world Anomaly Detection in Surveillance Videoscited by 577
  4. 2019Memorizing Normality to Detect Anomaly: Memory-augmented Deep Autoencoder for Unsupervised Anomaly Detectioncited by 564
  5. 2017Future Frame Prediction for Anomaly Detection -- A New Baselinecited by 391
  6. 2020Learning Memory-guided Normality for Anomaly Detectioncited by 278
  7. 2020Not only Look, but also Listen: Learning Multimodal Violence Detection under Weak Supervisioncited by 238
  8. 2021Weakly-supervised Video Anomaly Detection with Robust Temporal Feature Magnitude Learningcited by 230
  9. 2019Graph Convolutional Label Noise Cleaner: Train a Plug-and-play Action Classifier for Anomaly Detectioncited by 171
  10. 2020Anomaly Detection in Video via Self-Supervised and Multi-Task Learningcited by 148
  11. 2021MIST: Multiple Instance Self-Training Framework for Video Anomaly Detectioncited by 138
  12. 2023VadCLIP: Adapting Vision-Language Models for Weakly Supervised Video Anomaly Detectioncited by 123
38 more
  1. 2021Self-Supervised Predictive Convolutional Attentive Block for Anomaly Detectioncited by 119
  2. 2019Anomaly Detection in Video Sequence with Appearance-Motion Correspondencecited by 118
  3. 2018Object-centric Auto-encoders and Dummy Anomalies for Abnormal Event Detection in Videocited by 111
  4. 2019Learning Regularity in Skeleton Trajectories for Anomaly Detection in Videoscited by 104
  5. 2021UBnormal: New Benchmark for Supervised Open-Set Video Anomaly Detectioncited by 93
  6. 2023Dual Memory Units with Uncertainty Regulation for Weakly Supervised Video Anomaly Detectioncited by 90
  7. 2024Harnessing Large Language Models for Training-free Video Anomaly Detectioncited by 87
  8. 2020CLAWS: Clustering Assisted Weakly Supervised Learning with Normalcy Suppression for Anomalous Event Detectioncited by 79
  9. 2020Cloze Test Helps: Effective Video Anomaly Detection via Learning to Complete Video Eventscited by 70
  10. 2022CLIP-TSA: CLIP-Assisted Temporal Self-Attention for Weakly-Supervised Video Anomaly Detectioncited by 67
  11. 2023Unbiased Multiple Instance Learning for Weakly Supervised Video Anomaly Detectioncited by 64
  12. 2020Few-shot Scene-adaptive Anomaly Detectioncited by 62
  13. 2022Video Anomaly Detection by Solving Decoupled Spatio-Temporal Jigsaw Puzzlescited by 62
  14. 2023Open-Vocabulary Video Anomaly Detectioncited by 57
  15. 2024Follow the Rules: Reasoning for Video Anomaly Detection with Large Language Modelscited by 57
  16. 2023Video Event Restoration Based on Keyframes for Video Anomaly Detectioncited by 54
  17. 2020Localizing Anomalies from Weakly-Labeled Videoscited by 52
  18. 2024Text Prompt with Normality Guidance for Weakly Supervised Video Anomaly Detectioncited by 49
  19. 2022Exploiting Completeness and Uncertainty of Pseudo Labels for Weakly Supervised Video Anomaly Detectioncited by 47
  20. 2022SSMTL++: Revisiting Self-Supervised Multi-Task Learning for Video Anomaly Detectioncited by 45
  21. 2023Learning Prompt-Enhanced Context Features for Weakly-Supervised Video Anomaly Detectioncited by 43
  22. 2023A New Comprehensive Benchmark for Semi-supervised Video Anomaly Detection and Anticipationcited by 42
  23. 2019The Fairness of Risk Scores Beyond Classification: Bipartite Ranking and the xAUC Metriccited by 41
  24. 2024Weakly Supervised Video Anomaly Detection and Localization with Spatio-Temporal Promptscited by 40
  25. 2023Self-Distilled Masked Auto-Encoders are Efficient Video Anomaly Detectorscited by 37
  26. 2025EventVAD: Training-Free Event-Aware Video Anomaly Detectioncited by 36
  27. 2021Learning Not to Reconstruct Anomaliescited by 31
  28. 2021Synthetic Temporal Anomaly Guided End-to-End Video Anomaly Detectioncited by 26
  29. 2022Modality-Aware Contrastive Instance Learning with Self-Distillation for Weakly-Supervised Audio-Visual Violence Detectioncited by 26
  30. 2020Unsupervised Video Anomaly Detection via Normalizing Flows with Implicit Latent Featurescited by 23
  31. 2023BatchNorm-based Weakly Supervised Video Anomaly Detectioncited by 22
  32. 2024MULDE: Multiscale Log-Density Estimation via Denoising Score Matching for Video Anomaly Detectioncited by 22
  33. 2022An Attribute-based Method for Video Anomaly Detectioncited by 21
  34. 2025VADTree: Explainable Training-Free Video Anomaly Detection via Hierarchical Granularity-Aware Treecited by 18
  35. 2021FastAno: Fast Anomaly Detection via Spatio-temporal Patch Transformationcited by 15
  36. 2025AnyAnomaly: Zero-Shot Customizable Video Anomaly Detection with LVLMcited by 15
  37. 2024Video Anomaly Detection with Motion and Appearance Guided Patch Diffusion Modelcited by 14
  38. 2025A Unified Reasoning Framework for Holistic Zero-Shot Video Anomaly Analysiscited by 12

50 most cited of 77 All 77 (JSON)

Built on by

Not cited yet.