Research map / Graph learning / Knowledge and temporal graphs
Graph Convolutional Networks for Temporal Action Localization
Runhao Zeng, Wenbing Huang, Mingkui Tan, Yu Rong, Peilin Zhao, Junzhou Huang, Chuang Gan
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 |
|---|---|---|---|
| replace both GCNs with MLPs, i.e. remove proposal-proposal relations (reverse: add GCNs) (ablation) | P-GCN · THUMOS14 (RGB) | mAP@tIoU=0.5 (%): 34.75 -> 37.27 | Improved |
| add GCNs in place of MLPs (ablation) | P-GCN · THUMOS14 (Flow) | mAP@tIoU=0.5 (%): 43.68 -> 46.53 | Improved |
| use graph convolution instead of mean pooling (ablation) | P-GCN · THUMOS14 (RGB) | mAP@tIoU=0.5 (%): 35.32 -> 37.27 | Improved |
| use graph convolution instead of mean pooling (ablation) | P-GCN · THUMOS14 (Flow) | mAP@tIoU=0.5 (%): 43.97 -> 46.53 | Improved |
| mean pooling over adjacent proposals instead of independent MLP (ablation) | MLP baseline · THUMOS14 (RGB) | mAP@tIoU=0.5 (%): 34.75 -> 35.32 | Improved |
| add surrounding edges (ablation) | P-GCN · THUMOS14 (RGB) | mAP@tIoU=0.5 (%): 35.84 -> 37.27 | Improved |
| add contextual edges (ablation) | P-GCN · THUMOS14 (RGB) | mAP@tIoU=0.5 (%): 36.81 -> 37.27 | Improved |
| add contextual edges (ablation) | P-GCN · THUMOS14 (Flow) | mAP@tIoU=0.5 (%): 45.57 -> 46.53 | Improved |
| add self-addition of node feature (ablation) | P-GCN · THUMOS14 (RGB) | mAP@tIoU=0.5 (%): 35.55 -> 37.27 | Improved |
| add self-addition of node feature (ablation) | P-GCN · THUMOS14 (Flow) | mAP@tIoU=0.5 (%): 44.83 -> 46.53 | Improved |
| embed features before cosine similarity (ablation) | P-GCN · THUMOS14 (RGB) | mAP@tIoU=0.5 (%): 37.27 -> 36.81 | No clear effect |
| add boundary regression (ablation) | P-GCN · THUMOS14 (RGB) | mAP@tIoU=0.5 (%): 36.4 -> 37.3 | Improved |
Closest papers by meaning
- Graph Convolutional Module for Temporal Action Localization in Videos (TPAMI 2021 · similarity 0.95)
- G-TAD: Sub-Graph Localization for Temporal Action Detection (CVPR 2020 · similarity 0.93)
- ACGNet: Action Complement Graph Network for Weakly-Supervised Temporal Action Localization (AAAI 2022 · similarity 0.90)
- Temporal Action Localization in the Deep Learning Era: A Survey (TPAMI 2023 · similarity 0.90)
- Graph Attention Based Proposal 3D ConvNets for Action Detection (AAAI 2020 · similarity 0.90)
- Are Spatial-Temporal Graph Convolution Networks for Human Action Recognition Over-Parameterized? (CVPR 2025 · 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.
