Research map / Video / Action recognition
Adaptive Latent Trajectory Anchoring for Action Segmentation Dataset Condensation
Arthème Gauthier-Villars, Guodong Ding, Angela Yao
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 |
|---|---|---|---|
| replace GNI latent optimization with DDIM latent anchoring | TAS condensation (ASFormer) · Breakfast | Acc: 62.8 -> 68.0 | Improved |
| use diffusion latent anchoring instead of Coreset | TAS condensation (ASFormer) · Breakfast | Acc: 49.9 -> 68.0 | Improved |
| add adaptive anchor allocation to fixed anchors (ablation) | Ours · Breakfast (MSTCN) | Acc: 54.8 -> 63.4 | Improved |
| add adaptive anchor allocation to fixed anchors (ablation) | Ours · 50Salads (ASFormer) | Acc: 75.1 -> 71.2 | Hurt |
| add adaptive anchor allocation to fixed anchors (ablation) | Ours · 50Salads (ASFormer) | Edit: 64.9 -> 60.7 | Hurt |
| add adaptive anchor allocation at low budget (ablation) | Ours · GTEA (MSTCN), K=4 | Acc: 68.6 -> 70.5 | Improved |
| increase anchor budget from K=8 to K=10 (ablation) | Ours · GTEA (MSTCN) | Acc: 72.1 -> 73.3 | Improved |
| increase anchor budget from K=8 to K=10 (ablation) | Ours · GTEA (MSTCN) | storage ratio: 0.24 -> 0.29 | Hurt |
| increase fixed-anchor budget from K=2 to K=4 (ablation) | Ours-dagger · GTEA (MSTCN) | Acc: 64.4 -> 68.6 | Improved |
| increase GNI sub-segments from 8 to 64 (ablation) | GNI · GTEA (MSTCN) | Edit: 71.9 -> 66.9 | Hurt |
| increase GNI sub-segments from 8 to 64 (ablation) | GNI · GTEA (MSTCN) | Acc: 66.7 -> 66.6 | No clear effect |
| interpolate in latent space instead of feature space (ablation) | Ours · 50Salads | segmentation metrics | Improved |
Closest papers by meaning
- Condensing Action Segmentation Datasets via Generative Network Inversion (CVPR 2025 · similarity 0.91)
- Diffusion Action Segmentation (ICCV 2023 · similarity 0.89)
- DiffAct++: Diffusion Action Segmentation (TPAMI 2024 · similarity 0.88)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · similarity 0.87)
- Adaptive Residual Quantization for Memory-Efficient Temporal Action Segmentation (NeurIPS 2026 · similarity 0.87)
- 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.
