Research map / Video / Action recognition
OnPoint: Offline-to-Online Multi-Level Distillation for Point-Supervised Online Temporal Action Localization
Sakib Reza, Gauri Jagatap, Mohsen Moghaddam, OCTAVIA CAMPS, Andrea Fanelli
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 |
|---|---|---|---|
| add window-level anticipation distillation (ablation) | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: 60.0 -> 61.9 | Improved |
| add actionness-calibrated attention (ASAC) (ablation) | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: 59.9 -> 61.9 | Improved |
| add CASS distillation and calibration plus anticipation (ablation) | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: 57.5 -> 61.9 | Improved |
| add anchor-level point supervision (classification and regression) (ablation) | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: 58.4 -> 61.9 | Improved |
| add anchor-level point classification head (ablation) | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: 58.8 -> 61.9 | Improved |
| add anchor-level point regression head (ablation) | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: 59.5 -> 61.9 | Improved |
| use r+log(r) calibration instead of suppress-only, enhance-only, scaling or direct suppression (ablation) | ASAC · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: 61.5/61.3/58.3/59.8 -> 61.9 | Improved |
| replace ONMS with OSN post-processing (ablation) | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: 61.9 -> 51.6; AEDT -0.17 -> -1.53 s | Mixed |
| add our components to different offline teachers (HR-Pro, SMBD, LACP, TSASPC) (ablation) | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: TSASPC 52.9 -> 57.0; HR-Pro (RP) 57.9 -> 61.9 | Improved |
| distill from offline point-supervised teacher vs distillation-free anchor point supervision | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.5]: 33.3 -> 61.9 | Improved |
| OnPoint vs strongest point-supervised online baseline | OnPoint · THUMOS'14 | avg mAP@tIoU[0.1:0.7]: 46.3 -> 51.1 | Improved |
| OnPoint vs TSASPC+OAT-ONMS | OnPoint · EGTEA | avg mAP@tIoU[0.1:0.5]: 19.7 -> 23.1 | Improved |
Closest papers by meaning
- A Sliding Window Scheme for Online Temporal Action Localization (ECCV 2022 · similarity 0.92)
- Pointly-Supervised Action Localization (IJCV 2018 · similarity 0.90)
- Spot On: Action Localization from Pointly-Supervised Proposals (ECCV 2016 · similarity 0.90)
- Realigning Confidence with Temporal Saliency Information for Point-Level Weakly-Supervised Temporal Action Localization (CVPR 2024 · similarity 0.89)
- HR-Pro: Point-Supervised Temporal Action Localization via Hierarchical Reliability Propagation (AAAI 2024 · similarity 0.89)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.88)
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.
