Research map / Video / Action recognition
Causal Yet Future-Aware: Dual-Path Temporal Modeling for Online Action Segmentation
Zhichao Zheng, Peirong Ma, Ying Zhou, Li Kong, Junsheng Zhou
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 |
|---|---|---|---|
| add predictive distillation to temporal decomposition (ablation) | DPTM (MS-TCN) · 50Salads | Avg: 39.0 -> 48.3 | Improved |
| add temporal decomposition to predictive distillation (ablation) | DPTM (MS-TCN) · 50Salads | Avg: 43.5 -> 48.3 | Improved |
| use predictive distillation instead of direct distillation (ablation) | DPTM (MS-TCN) · Breakfast | Avg: 26.5 -> 30.5 | Improved |
| use disentangled predictive distillation instead of frame-wise future prediction (ablation) | DPTM (MS-TCN) · Breakfast | Avg: 26.9 -> 30.5 | Improved |
| use linear dilation instead of fixed dilation in future estimator (ablation) | DPTM (MS-TCN) · GTEA | Avg: 63.0 -> 72.7 | Improved |
| use linear dilation instead of exponential dilation in future estimator (ablation) | DPTM (MS-TCN) · GTEA | Avg: 67.3 -> 72.7 | Improved |
| replace offline bidirectional MS-TCN with dual-path decomposition (MS-TCN+DS) (ablation) | MS-TCN · GTEA | Avg: 78.9 -> 81.0 | No clear effect |
| replace offline bidirectional MS-TCN with dual-path decomposition (MS-TCN+DS) (ablation) | MS-TCN · 50Salads | Avg: 72.7 -> 73.5 | No clear effect |
| scale MS-TCN encoder to 1.2M parameters (ablation) | MS-TCN · 50Salads | Avg: 72.7 -> 73.6 | No clear effect |
| increase embedding dimension from 32 to 128 (ablation) | DPTM (MS-TCN) · 50Salads | Avg: 42.3 -> 50.2 | Improved |
| increase embedding dimension from 128 to 256 (ablation) | DPTM (MS-TCN) · 50Salads | Avg: 50.2 -> 48.5 | No clear effect |
| apply DPTM to MS-TCN online baseline | MS-TCN · 50Salads | Avg: 37.2 -> 48.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- OnlineTAS — OnlineTAS: An Online Baseline for Temporal Action Segmentation (NeurIPS 2024) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- OnlineTAS: An Online Baseline for Temporal Action Segmentation (NeurIPS 2024 · similarity 0.90)
- Two-Stage Active Learning for Efficient Temporal Action Segmentation (ECCV 2024 · similarity 0.88)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.87)
- Towards Understanding Future: Consistency Guided Probabilistic Modeling for Action Anticipation (AAAI 2024 · similarity 0.86)
- AntGPT: Can Large Language Models Help Long-term Action Anticipation from Videos? (ICLR 2024 · similarity 0.86)
- Progress-Aware Online Action Segmentation for Egocentric Procedural Task Videos (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.
