Research map / Video / Action recognition
Pointly-Supervised Action Localization
Pascal Mettes, Cees G. M. Snoek
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 |
|---|---|---|---|
| Use point supervision instead of box supervision | Spatio-temporal proposal mining · UCF-101 | mAP@0.2: 0.399 -> 0.393 | No clear effect |
| Use point supervision instead of video-label only supervision | Proposal mining · UCF Sports and UCF-101 | localization | Improved |
| Annotate a point every 100 frames instead of every frame (ablation) | Point supervision · UCF-101 | mAP@0.2: 0.393 -> 0.371 | No clear effect |
| Annotate a point every 5 frames instead of every frame (ablation) | Point supervision · UCF-101 | mAP@0.2: 0.393 -> 0.389 | No clear effect |
| Remove 50% of low-quality proposals (oracle ground truth added) (ablation) | Pointly-supervised localization · UCF Sports | localization performance: 0.23 -> 0.32 | Improved |
| Remove 95% of low-quality proposals (oracle ground truth added) (ablation) | Pointly-supervised localization · UCF Sports | localization performance: 0.23 -> 0.49 | Improved |
| Keep only ground truth tube and high-overlap proposals (ablation) | Pointly-supervised localization · UCF Sports | localization performance: 0.23 -> 0.90 | Improved |
| Add Gaussian noise to point annotations (sigma up to 5) (ablation) | Point supervision · UCF-101 | localization | No clear effect |
| Add large Gaussian noise to point annotations (sigma 50 or 100) (ablation) | Point supervision · UCF-101 | localization at IoU 0.1-0.4 | Hurt |
| Add person-detection pseudo-points at inference, IoU 0.5 (ablation) | Pointly-supervised localization · UCF Sports | relative localization performance: +20.0% | Improved |
| Add independent-motion pseudo-points at inference, IoU 0.5 (ablation) | Pointly-supervised localization · UCF Sports | relative localization performance: +6.7% | Improved |
| Add self-supervision pseudo-points at inference, IoU 0.5 (ablation) | Pointly-supervised localization · UCF Sports | relative localization performance: +6.8% | Improved |
Closest papers by meaning
- Spot On: Action Localization from Pointly-Supervised Proposals (ECCV 2016 · similarity 0.96)
- Temporal Action Localization in the Deep Learning Era: A Survey (TPAMI 2023 · similarity 0.93)
- HR-Pro: Point-Supervised Temporal Action Localization via Hierarchical Reliability Propagation (AAAI 2024 · similarity 0.90)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · similarity 0.90)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.90)
- Realigning Confidence with Temporal Saliency Information for Point-Level Weakly-Supervised Temporal Action Localization (CVPR 2024 · similarity 0.90)
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.
