Research map / Video / Action recognition
DeST: A Decoupled Spatio-Temporal Framework for Action Segmentation
Yunheng Li, Zhong-Yu Li, Shanghua Gao, Qilong Wang, Qibin Hou, Ming-Ming Cheng
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 joint-decoupled temporal modeling (JTM) instead of joint-shared 2D-Conv (ablation) | DeST with TCN · MCFS-130 | F1@50: 3.0% | Improved |
| use joint-decoupled JTM instead of joint-shared weight JWTM with linear transformer (ablation) | DeST · MCFS-22 | all metrics | Improved |
| use spatial sub-features instead of the same spatial features for every layer (ablation) | DSTI module · MCFS | performance | Improved |
| use cross-attention instead of summation for spatio-temporal interaction (ablation) | DSTI module · MCFS | performance | Improved |
| use Convolution or Avgpool instead of Maxpool as feature transformation (ablation) | DeST · MCFS | performance | Improved |
| increase number of JTM layers from 8 to 10 (ablation) | DeST · MCFS | performance: 8 -> 10 | Improved |
| increase number of JTM layers from 10 to 11 or 12 (ablation) | DeST · MCFS | performance: 10 -> 11 or 12 | Improved |
| increase number of DSTI layers Lc from 1 to 9 (ablation) | DeST · MCFS | performance: 1 -> 9 | Improved |
| add the TSR branch to DeST (DeST*) | DeST · MCFS-22 and MCFS-130 | segmentation scores | Improved |
Closest papers by meaning
- Segmental Spatiotemporal CNNs for Fine-Grained Action Segmentation (ECCV 2016 · similarity 0.88)
- Scene-Aware Spatiotemporal Generalization: Towards Robust Temporal Action Detection Across Domains (AAAI 2026 · similarity 0.87)
- Action Segmentation With Joint Self-Supervised Temporal Domain Adaptation (CVPR 2020 · similarity 0.87)
- OnlineTAS: An Online Baseline for Temporal Action Segmentation (NeurIPS 2024 · similarity 0.87)
- Building a Multi-modal Spatiotemporal Expert for Zero-shot Action Recognition with CLIP (AAAI 2025 · similarity 0.86)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 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.
