Research map / Video / Action recognition
ASPnet: Action Segmentation With Shared-Private Representation of Multiple Data Sources
Beatrice van Amsterdam, Abdolrahim Kadkhodamohammadi, Imanol Luengo, Danail Stoyanov
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 |
|---|---|---|---|
| Fuse video and accelerometer features with shared-private representations and a temporal attention bottleneck | ASFormer · 50Salads | F1@10: 84.8 -> 89.2 | Improved |
| Add accelerometer input to Br-Prompt RGB and optical flow | Br-Prompt+ASPnet · 50Salads | F1@10: 91.2 -> 92.7 | Improved |
| Use ASPnet with RGB-flow inputs instead of ASFormer feature concatenation | ASFormer · 50Salads | F1@10: 87.8 -> 88.6 | Improved |
| Use ASPnet instead of ASFormer with the same video input features | ASFormer · Breakfast | F1@10: 76.0 -> 78.1 | Improved |
| Use ASPnet instead of MA-TCN for video-based segmentation | MA-TCN · RARP45 | F1@10: 83.7 -> 86.7 | Improved |
| Add moving-average post-processing (ablation) | ASPnet · 50Salads | F1@10: 89.2 -> 90.4 | Improved |
| Initialize the attention bottleneck with learned shared features rather than a Gaussian (ablation) | ASPnet · 50Salads | F1@10: 86.5 -> 89.2 | Improved |
| Use MMD instead of Euclidean distance to align shared features (ablation) | ASPnet · 50Salads | F1@10: 87.0 -> 89.2 | Improved |
| Add private features to the prediction by removing the private-feature mask (ablation) | ASPnet · 50Salads | F1@10: 83.0 -> 90.4 | Improved |
| Add shared bottleneck features to the prediction by removing the shared-feature mask (ablation) | ASPnet · 50Salads | F1@10: 86.5 -> 90.4 | Improved |
| Share weights across encoder streams (ablation) | ASPnet · 50Salads | F1@10: 91.3 -> 91.0 | No clear effect |
| Use scalable MMD against the average bottleneck with shared encoder weights (ablation) | ASPnet · 50Salads | F1@10: 91.0 -> 90.9 | No clear effect |
Closest papers by meaning
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.88)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.88)
- HopaDIFF: Holistic-Partial Aware Fourier Conditioned Diffusion for Referring Human Action Segmentation in Multi-Person Scenarios (NeurIPS 2025 · similarity 0.86)
- Understanding Multi-Task Activities from Single-Task Videos (CVPR 2025 · similarity 0.86)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · similarity 0.86)
- OnlineTAS: An Online Baseline for Temporal Action Segmentation (NeurIPS 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.
