Research map / Video / Action recognition
Video Self-Stitching Graph Network for Temporal Action Localization
Chen Zhao, Ali K. Thabet, Bernard Ghanem
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 video self-stitching (VSS) to baseline (ablation) | VSGN baseline · THUMOS-14 | short-action mAP: 45.1 -> 48.7 | Improved |
| add VSS to baseline (ablation) | VSGN baseline · THUMOS-14 | mAP@0.5: 42.54 -> 45.52 | Improved |
| add xGPN to baseline (ablation) | VSGN baseline · THUMOS-14 | mAP@0.5: 42.54 -> 50.24 | Improved |
| add VSS on top of xGPN (ablation) | VSGN · THUMOS-14 | mAP@0.5: 50.24 -> 52.45 | Improved |
| add VSS on top of xGPN (ablation) | VSGN · THUMOS-14 | short-action mAP: 53.4 -> 56.6 | Improved |
| add VSS to baseline (ablation) | VSGN baseline · ActivityNet-v1.3 | average mAP: 34.25 -> 34.70 | Improved |
| add xGPN to baseline (ablation) | VSGN baseline · ActivityNet-v1.3 | average mAP: 34.25 -> 33.79 | Hurt |
| add xGPN on top of VSS (ablation) | VSGN · ActivityNet-v1.3 | average mAP: 34.70 -> 35.07 | Improved |
| combine Clip O and Clip U predictions (ablation) | VSGN · ActivityNet-v1.3 | average mAP: 34.96 -> 35.07 | Improved |
| use xGN at all encoder levels instead of a single level (ablation) | xGPN · ActivityNet-v1.3 | average mAP: 34.46 -> 35.07 | Improved |
| use half free and half cross-scale edges instead of only free edges (ablation) | xGN · ActivityNet-v1.3 | average mAP: 34.48 -> 35.07 | Improved |
| use half free and half cross-scale edges instead of only cross-scale edges (ablation) | xGN · ActivityNet-v1.3 | average mAP: 34.75 -> 35.07 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- G-TAD — G-TAD: Sub-Graph Localization for Temporal Action Detection (CVPR 2020) · better on 4 of 4 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
- Graph Convolutional Module for Temporal Action Localization in Videos (TPAMI 2021 · similarity 0.93)
- G-TAD: Sub-Graph Localization for Temporal Action Detection (CVPR 2020 · similarity 0.91)
- One-Shot Action Localization by Learning Sequence Matching Network (CVPR 2018 · similarity 0.90)
- Temporal Action Localization in the Deep Learning Era: A Survey (TPAMI 2023 · similarity 0.90)
- ACGNet: Action Complement Graph Network for Weakly-Supervised Temporal Action Localization (AAAI 2022 · similarity 0.90)
- Temporal Action Localization with Cross Layer Task Decoupling and Refinement (AAAI 2025 · similarity 0.89)
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.
