弱监督视频异常检测评估常误判，真实性能可能远低于报告值。
Frame-Level Evaluation in Weakly Supervised Video Anomaly Detection Mostly Measures Video-Level Ranking
- 提出分解Micro-AUROC为视频内与跨视频两部分，揭示评估偏差来源。
- 实验显示99%以上异常帧对比发生在不同视频间，模型可凭视频级区分获高分。
- 即使所有帧得分相同，仍能保持98.6%的评估分数，适合评估者与研究者参考。
弱监督视频异常检测器使用视频级标签训练，却常以帧级定位性能评估，采用对池化测试帧的Micro-AUROC或AP指标。由于这些指标比较不同视频的帧，检测器只需区分视频即可获得高分，无需精确排序视频内部时刻。我们精确分解Micro-AUROC，按视频身份分离出视频内（Within-AUROC）和跨视频（Cross-AUROC）两部分。在ShanghaiTech、XD-Violence和UCF-Crime数据集上，异常帧与正常帧之间的比较仅有0.071%-0.388%发生于同一视频内。当两类样本分布于V个视频时，该比例随1/V下降，此现象称为时间稀释（temporal dilution）。我们训练视频级异常分类器，在视频内重复输出相同分数。这些视频常数输出仍达到81.40%-97.18%的Micro-AUROC，且在72次控制实验中，用视频均值替代每帧得分，仅损失中位数1.4%的评估分数。该模式同样适用于作者公开结果及XD-Violence官方AP评估。表明检测器可在无视频内差异的情况下获得高分。
原文摘要 · Abstract (English)
Weakly supervised video anomaly detectors are trained with video-level labels but are commonly evaluated as temporal localizers using Micro-AUROC or AP over pooled test frames. Because these metrics compare frames from different videos, a detector can score well by separating videos without accurately ordering moments within them. We exactly decompose Micro-AUROC by video identity into Within-AUROC for temporal ordering within videos and Cross-AUROC for comparisons across videos. Across ShanghaiTech, XD-Violence, and UCF-Crime, only 0.071-0.388% of comparisons between anomalous and normal frames occur within the same video. When both classes remain distributed across V videos, this share decreases as O(1/V), a benchmark property we call temporal dilution. We train anomaly video binary classifiers under the same video-level supervision and repeat each video score across all frames. These video-constant outputs reach 81.40-97.18 Micro-AUROC despite having no within-video variation. Across 72 controlled runs, replacing every frame score with its video mean preserves a median 98.6% of the Micro-AUROC margin above chance. The same empirical pattern holds for author-released outputs and for XD-Violence under its official AP evaluation. A detector can therefore achieve a high pooled score even when it assigns the same score to every moment within each video.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。