Research map / Video / Action recognition
MS-TCT: Multi-Scale Temporal ConvTransformer for Action Detection
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 |
|---|---|---|---|
| Use the multi-scale ConvTransformer with center-relative heat-map supervision | I3D-based action detection · Charades | per-frame mAP: 15.6 -> 25.4 | Helped |
| Add the Temporal Encoder (ablation) | I3D feature baseline · Charades | per-frame mAP: 15.6 -> 23.6; + 7.0% | Helped |
| Add the TS Mixer (ablation) | Temporal Encoder · Charades | per-frame mAP: 23.6 -> 24.1; + 0.5% | Helped |
| Add the heat-map branch jointly with the classification branch (ablation) | MS-TCT · Charades | per-frame mAP: 24.1 -> 25.4 | Helped |
| Remove the classification branch (ablation) | MS-TCT · Charades | per-frame mAP: 25.4 -> 10.7 | Hurt |
| Remove the Global Relational Layer (ablation) | MS-TCT stage · Charades | per-frame mAP: 25.4 -> 20.9 | Hurt |
| Remove the Local Relational Layer (ablation) | MS-TCT stage · Charades | per-frame mAP: 25.4 -> 24.0 | Hurt |
| Replace Temporal Merge with stride-1 temporal convolution (ablation) | MS-TCT stage · Charades | per-frame mAP: 25.4 -> 22.7 | Hurt |
| Replace ConvTransformer stages with pure Transformer stages (ablation) | MS-TCT · Charades | mAP: 25.4 -> 22.3; + 3.1% | Hurt |
| Replace ConvTransformer stages with pure convolution stages (ablation) | MS-TCT · Charades | mAP: 25.4 -> 21.4; + 4.0% | Hurt |
| Add learnable positional embedding before the Temporal Encoder (ablation) | MS-TCT · Charades | mAP: -0.2% | Hurt |
| Increase heat-map Gaussian sigma from 1/8 to 1/2 of instance duration (ablation) | MS-TCT heat-map branch · Charades | mAP: 24.6 -> 25.4 | Helped |
Closest papers by meaning
- MS-TCN: Multi-Stage Temporal Convolutional Network for Action Segmentation (CVPR 2019 · similarity 0.90)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.90)
- Temporal Context Aggregation Network for Temporal Action Proposal Refinement (CVPR 2021 · similarity 0.88)
- Adapting Short-Term Transformers for Action Detection in Untrimmed Videos (CVPR 2024 · similarity 0.88)
- DiGIT: Multi-Dilated Gated Encoder and Central-Adjacent Region Integrated Decoder for Temporal Action Detection Transformer (CVPR 2025 · similarity 0.88)
- TIM: A Time Interval Machine for Audio-Visual Action Recognition (CVPR 2024 · similarity 0.86)
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.