Research map / Label-efficient and robust learning / Domain adaptation and generalization
Source-Free Video Domain Adaptation by Learning Temporal Consistency for Action Recognition
Yuecong Xu, Jianfei Yang, Haozhi Cao, Keyu Wu, Min Wu, Zhenghua Chen
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 source prediction consistency to feature consistency (FC only -> full) (ablation) | ATCoN · UCF-HMDB-full | top-1 accuracy: 77.78 -> 79.72 | Improved |
| add feature consistency to source prediction consistency (ablation) | ATCoN · UCF-HMDB-full | top-1 accuracy: 77.50 -> 78.89 | Improved |
| extend source prediction consistency to the overall target prediction (ablation) | ATCoN · UCF-HMDB-full | top-1 accuracy: 76.67 -> 77.50 | Improved |
| add feature consistency plus source prediction consistency to source-only (ablation) | ATCoN · UCF-HMDB-full | top-1 accuracy: 72.78 -> 78.89 | Improved |
| add IM loss and pseudo-labeling on top of temporal consistency (ablation) | ATCoN-TC · UCF-HMDB-full | top-1 accuracy: 78.89 -> 79.72 | Improved |
| add IM loss and pseudo-labeling on top of temporal consistency (H51->U101) (ablation) | ATCoN-TC · UCF-HMDB-full | top-1 accuracy: 84.59 -> 85.29 | Improved |
| add local weight module (any form) (ablation) | ATCoN-NA · UCF-HMDB-full | top-1 accuracy: 78.33 -> 79.72 | Improved |
| apply local relevance weight to both feature and prediction (vs feature only) (ablation) | ATCoN-A@F · UCF-HMDB-full | top-1 accuracy: 79.17 -> 79.72 | Improved |
| apply local relevance weight to both feature and prediction (vs prediction only) (ablation) | ATCoN-A@P · UCF-HMDB-full | top-1 accuracy: 84.41 -> 85.29 | Improved |
| add local weight module (H51->U101) (ablation) | ATCoN-NA · UCF-HMDB-full | top-1 accuracy: 83.89 -> 85.29 | Improved |
| use ATCoN instead of image-based SFDA methods (SHOT, MA, CPGA) | source-free video adaptation · Daily-DA | top-1 accuracy: 26.46 -> 33.53 | Improved |
| apply image-based SFDA method SHOT instead of source-only | TRN · Daily-DA | top-1 accuracy: 27.89 -> 25.30 | Hurt |
Closest papers by meaning
- Source-Free Video Domain Adaptation With Spatial-Temporal-Historical Consistency Learning (CVPR 2023 · similarity 0.93)
- Unsupervised Video Domain Adaptation for Action Recognition: A Disentanglement Perspective (NeurIPS 2023 · similarity 0.89)
- The Unreasonable Effectiveness of Large Language-Vision Models for Source-Free Video Domain Adaptation (ICCV 2023 · similarity 0.89)
- Modality-Collaborative Test-Time Adaptation for Action Recognition (CVPR 2024 · similarity 0.89)
- Understanding and Improving Source-free Domain Adaptation from a Theoretical Perspective (CVPR 2024 · similarity 0.87)
- Unsupervised Video Domain Adaptation with Masked Pre-Training and Collaborative Self-Training (CVPR 2024 · similarity 0.87)
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.
