Research map / Video / Action recognition
C2F-TCN: A Framework for Semi- and Fully-Supervised Temporal Action Segmentation
Dipika Singhania, Rahul Rahaman, 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 |
|---|---|---|---|
| add coarse-to-fine ensemble of decoder outputs (ablation) | base model M · Breakfast | F1@10: +5.1 F1@10, +3.1 Edit | Improved |
| add coarse-to-fine ensemble (ablation) | ED-TCN · 50Salads | MoF: 62.7 -> 63.3 Edit; 66.5 -> 68.5 MoF | Improved |
| add feature augmentation (FA-Train-Test) (ablation) | C2F-TCN · 50Salads | MoF: 79.6 -> 84.5 | Improved |
| add feature augmentation (FA-Train-Test) (ablation) | C2F-TCN · Breakfast | MoF: 70.2 -> 76.6 | Improved |
| add feature augmentation (FA-Train-Test) (ablation) | MS-TCN · Breakfast | MoF: 67.6 -> 71.3 | Improved |
| add feature augmentation (ablation) | MS-TCN · GTEA | F1@25: F1@25 86.2 -> 85.7; F1@50 74.4 -> 74.1 | Mixed |
| add temporal pyramid pooling layer (ablation) | C2F-TCN + FA · 50Salads | F1@50: 71.0 -> 72.8 | Improved |
| add loss at each decoder layer instead of ensemble (ablation) | base model M · Breakfast | Edit: Edit 60.1 -> 58.7; MoF 69.3 -> 70.6 | Hurt |
| use learned ensemble weights instead of fixed (ablation) | C2F-TCN · Breakfast | Edit: Edit 63.2 -> 61.4; MoF 70.2 -> 70.5 | Hurt |
| use fixed equal ensemble weights (ablation) | base model M · Breakfast | Edit: 60.1 -> 63.2 | Improved |
| add time-proximity condition to contrastive learning (ablation) | unsupervised C2F-TCN · Breakfast | MoF: 36.1 -> 50.0 | Improved |
| add video-level constraint to contrastive learning (ablation) | unsupervised C2F-TCN · Breakfast | MoF: 50.0 -> 66.1 | Improved |
Later papers that compared with this method
2 later accepted papers report C2F-TCN as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- ViSTec: Video Modeling for Sports Technique Recognition and Tactical Analysis (AAAI 2024 · better on 5 of 5 shared results)
- How Much Temporal Long-Term Context is Needed for Action Segmentation? (ICCV 2023 · better on 1 of 2 shared results)
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.
- MS-TCN — MS-TCN: Multi-Stage Temporal Convolutional Network for Action Segmentation (CVPR 2019) · better on 4 of 4 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
- Iterative Contrast-Classify for Semi-supervised Temporal Action Segmentation (AAAI 2022 · similarity 0.93)
- MS-TCN: Multi-Stage Temporal Convolutional Network for Action Segmentation (CVPR 2019 · similarity 0.91)
- Leveraging Action Affinity and Continuity for Semi-Supervised Temporal Action Segmentation (ECCV 2022 · similarity 0.91)
- SCT: Set Constrained Temporal Transformer for Set Supervised Action Segmentation (CVPR 2020 · similarity 0.90)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.89)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.89)
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.
