Research map / Video / Action recognition
Temporal Relational Modeling with Self-Supervision for Action Segmentation
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 |
|---|---|---|---|
| Add DTGRM to the MS-TCN baseline | MS-TCN · 50Salads | Edit: 67.9 -> 72.0 | Helped |
| Add DTGRM to the MS-TCN baseline | MS-TCN · GTEA | Edit: 79.0 -> 83.0 | Helped |
| Add DTGRM to the MS-TCN baseline | MS-TCN · Breakfast | Edit: 61.7 -> 68.9 | Helped |
| Replace DRGC graph convolutions with dilated temporal convolution (ablation) | DTGRM · GTEA | F1@10: 86.3 -> 87.3 | Helped |
| Add S-Graph to the L-Graph-only model (ablation) | DTGRM without self-supervision · GTEA | F1@10: 85.6 -> 87.3 | Helped |
| Add L-Graph to the S-Graph-only model (ablation) | DTGRM without self-supervision · GTEA | F1@10: 48.2 -> 87.3 | Helped |
| Add auxiliary self-supervision (ablation) | DTGRM · 50Salads | F1@10: 74.0 -> 79.1 | Helped |
| Add one DTGRM stage to the backbone (ablation) | MS-TCN backbone · 50Salads | F1@10: 52.7 -> 69.0 | Helped |
| Increase the number of DTGRM stages from one to three (ablation) | MS-TCN backbone · 50Salads | F1@10: 69.0 -> 79.1 | Helped |
| Use only S-Graph, omitting L-Graph (ablation) | DTGRM without self-supervision · GTEA | F1@10: 87.3 -> 48.2 | Hurt |
| Use only L-Graph, omitting S-Graph (ablation) | DTGRM without self-supervision · GTEA | F1@10: 87.3 -> 85.6 | Hurt |
| Set λc to 0, removing the correction loss (ablation) | DTGRM self-supervision · 50Salads | F1@10: 79.1 -> 58.7 | Hurt |
Closest papers by meaning
- Improving Action Segmentation via Graph-Based Temporal Reasoning (CVPR 2020 · similarity 0.92)
- Object-Relation Reasoning Graph for Action Recognition (CVPR 2022 · similarity 0.88)
- Graph-Based High-Order Relation Modeling for Long-Term Action Recognition (CVPR 2021 · similarity 0.88)
- Temporal Convolutional Networks for Action Segmentation and Detection (CVPR 2017 · similarity 0.88)
- G-TAD: Sub-Graph Localization for Temporal Action Detection (CVPR 2020 · similarity 0.88)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (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.