Research map / Video / Action recognition
TriDet: Temporal Action Detection With Relative Boundary Modeling
Read the paperCheck 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 |
|---|---|---|---|
| use relative boundary distributions with Trident-head instead of instant-level boundary regression (ablation) | TriDet with SGP · THUMOS14 | average mAP: 68.3 -> 69.3 | Helped |
| replace self-attention with SGP (ablation) | TriDet feature pyramid · THUMOS14 | average mAP: 66.8 -> 68.3 | Helped |
| replace the two-layer convolutional baseline pyramid with SGP (ablation) | TriDet feature pyramid · THUMOS14 | average mAP: 62.1 -> 68.3 | Helped |
| replace self-attention with a 1D convolution using the same receptive field (ablation) | Transformer baseline · THUMOS14 | average mAP: 66.8 -> 64.9 | Hurt |
| replace Layer Normalization with Group Normalization (ablation) | Transformer baseline · THUMOS14 | average mAP: 66.8 -> 64.8 | Hurt |
| use a hybrid of Layer Normalization and Group Normalization (ablation) | Transformer baseline · THUMOS14 | average mAP: 66.8 -> 65.7 | Hurt |
| add Trident-head to LN-GN-Mix (ablation) | LN-GN-Mix convolutional structure · THUMOS14 | average mAP: 65.7 -> 66.0 | Helped |
| use SlowFast instead of I3D features | TriDet · HACS-Segment | average mAP: 36.8 -> 38.6 | Helped |
| increase the number of boundary bins from 4 to 14 (ablation) | TriDet · HACS-Segment | average mAP: 33.3 -> 38.6 | Helped |
| increase the number of boundary bins from 4 to 16 (ablation) | TriDet · THUMOS14 | average mAP: 68.1 -> 69.3 | Helped |
| increase feature pyramid levels from 1 to 6 (ablation) | TriDet · THUMOS14 | average mAP: 44.5 -> 69.3 | Helped |
| increase feature pyramid levels from 6 to 7 (ablation) | TriDet · THUMOS14 | average mAP: 69.3 -> 68.9 | Hurt |
Closest papers by meaning
- Dual DETRs for Multi-Label Temporal Action Detection (CVPR 2024 · similarity 0.86)
- Temporal Action Localization with Cross Layer Task Decoupling and Refinement (AAAI 2025 · similarity 0.85)
- DiGIT: Multi-Dilated Gated Encoder and Central-Adjacent Region Integrated Decoder for Temporal Action Detection Transformer (CVPR 2025 · similarity 0.85)
- TIM: A Time Interval Machine for Audio-Visual Action Recognition (CVPR 2024 · similarity 0.85)
- TE-TAD: Towards Full End-to-End Temporal Action Detection via Time-Aligned Coordinate Expression (CVPR 2024 · similarity 0.85)
- End-to-End Temporal Action Detection with 1B Parameters Across 1000 Frames (CVPR 2024 · similarity 0.85)
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.