Research map / Video / Action recognition
Gaussian Temporal Awareness Networks for Action Localization
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 fixed temporal scales with learned Gaussian kernels and Gaussian pooling (ablation) | One-stage action localization · THUMOS14 | mAP: 33.5 to 37.1 | Helped |
| Replace fixed temporal scales with learned Gaussian kernels and Gaussian pooling (ablation) | One-stage action localization · ActivityNet v1.3 | mAP: 29.8 to 31.6 | Helped |
| Add Gaussian grouping to merge overlapping kernels (ablation) | GTAN · THUMOS14 | mAP: 37.1 to 38.2 | Helped |
| Add Gaussian grouping to merge overlapping kernels (ablation) | GTAN · ActivityNet v1.3 | mAP: 31.6 to 34.3 | Helped |
| Add Gaussian grouping (ablation) | GTAN · THUMOS14 actions ≥ 128 frames | mAP: 22.1 to 25.9 | Helped |
| Add Gaussian grouping (ablation) | GTAN · ActivityNet v1.3 actions ≥ 2048 frames | mAP: 49.4 to 54.2 | Helped |
| Use 4 anchor layers rather than SSAD's fixed-scale anchors (ablation) | Action localization · ActivityNet v1.3 | average mAP: 26.77% in average mAP; SSAD 5.12% | Helped |
| Use GTAN instead of BSN | Temporal action detection · THUMOS14 test | mAP@0.5: 36.9 to 38.8 | Helped |
| Use GTAN instead of BSN | Temporal action detection · ActivityNet v1.3 validation | average mAP: 33.72 to 34.31 | Helped |
| Use GTAN instead of BSN | Temporal action detection · ActivityNet v1.3 testing | average mAP: 34.42 to 35.54 | Helped |
Closest papers by meaning
- G-TAD: Sub-Graph Localization for Temporal Action Detection (CVPR 2020 · similarity 0.90)
- CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos (CVPR 2017 · similarity 0.90)
- Temporal Action Localization by Structured Maximal Sums (CVPR 2017 · similarity 0.90)
- Temporal Action Localization in Untrimmed Videos via Multi-Stage CNNs (CVPR 2016 · similarity 0.89)
- Rethinking the Faster R-CNN Architecture for Temporal Action Localization (CVPR 2018 · similarity 0.89)
- Temporal Action Localization with Cross Layer Task Decoupling and Refinement (AAAI 2025 · similarity 0.87)
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.