Research map / Video / Video segmentation and tracking
Sequential keypoint density estimator: an overlooked baseline of skeleton-based video anomaly detection
Anja Delić, Matej Grcic, Siniša Šegvić
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| learn diagonal covariance instead of fixed identity (ablation) | SeeKer · UBnormal (full) | AUROC: 61.4 -> 77.9 | Improved |
| learn diagonal covariance instead of fixed identity (ablation) | SeeKer · ShanghaiTech (full) | AUROC: 74.1 -> 85.5 | Improved |
| learned full covariance instead of diagonal (ablation) | SeeKer · ShanghaiTech (full) | AUROC: 85.5 -> 84.3 | No clear effect |
| keypoint-level instead of skeleton-level factorization (ablation) | SeeKer · UBnormal (full) | AUROC: 75.0 -> 77.9 | Improved |
| keypoint-level instead of skeleton-level factorization (ablation) | SeeKer · ShanghaiTech (full) | AUROC: 79.0 -> 85.5 | Improved |
| weight keypoint log-densities by detector confidence (ablation) | SeeKer · UBnormal (full) | AUROC: 76.8 -> 77.9 | Improved |
| weight keypoint log-densities by detector confidence (ablation) | SeeKer · ShanghaiTech (full) | AUROC: 83.7 -> 85.5 | Improved |
| train only on skeletons with confidence at least 0.4 plus weighting (ablation) | SeeKer · UBnormal (full) | AUROC: 75.5 -> 77.9 | Improved |
| add Gaussian smoothing (ablation) | SeeKer · UBnormal (full) | AUROC: 76.8 -> 77.9 | Improved |
| add Gaussian smoothing (ablation) | SeeKer · ShanghaiTech (full) | AUROC: 83.4 -> 85.5 | Improved |
| add Gaussian smoothing (ablation) | STG-NF · UBnormal (full) | AUROC: 68.2 -> 71.8 | Improved |
| masked fully connected model instead of transformer decoder (ablation) | SeeKer · UBnormal (full) | AUROC: 71.1 -> 77.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- MULDE — MULDE: Multiscale Log-Density Estimation via Denoising Score Matching for Video Anomaly Detection (CVPR 2024) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- STEP: Score-Based Temporal Energy for Human Pose Video Anomaly Detection (ECCV 2026 · similarity 0.90)
- Learning Regularity in Skeleton Trajectories for Anomaly Detection in Videos (CVPR 2019 · similarity 0.88)
- Multimodal Motion Conditioned Diffusion Model for Skeleton-based Video Anomaly Detection (ICCV 2023 · similarity 0.88)
- Dual Conditioned Motion Diffusion for Pose-Based Video Anomaly Detection (AAAI 2025 · similarity 0.85)
- Video Anomaly Detection with Motion and Appearance Guided Patch Diffusion Model (AAAI 2025 · similarity 0.84)
- Holmes-VAU: Towards Long-term Video Anomaly Understanding at Any Granularity (CVPR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
