Research map / Label-efficient and robust learning / Domain adaptation and generalization
Temporal Attentive Alignment for Large-Scale Video Domain Adaptation
Min-Hung Chen, Zsolt Kira, Ghassan AlRegib, Jaekwon Yoo, Ruxin Chen, Jian Zheng
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 domain attention to the TA2N model (all discriminators) (ablation) | TA2N (TemRelation) · UCF-HMDB full U->H | accuracy: 77.22 -> 78.33 | Improved |
| add domain attention to the TA2N model (all discriminators) (ablation) | TA2N (TemRelation) · UCF-HMDB full H->U | accuracy: 80.56 -> 81.79 | Improved |
| use domain attention instead of general attention (ablation) | TA3N · UCF-HMDB full H->U | accuracy: 80.91 -> 81.79 | Improved |
| use domain attention instead of general attention (ablation) | TA3N · UCF-HMDB full U->H | accuracy: 77.22 -> 78.33 | Improved |
| add domain attention to TemPooling (ablation) | TemPooling with all discriminators · UCF-HMDB full H->U | accuracy: 75.13 -> 78.46 | Improved |
| add domain attention to TemPooling (ablation) | TemPooling with all discriminators · UCF-HMDB full U->H | accuracy: 71.11 -> 73.06 | Improved |
| add relation discriminator and combine all three discriminators (ablation) | TemRelation · UCF-HMDB full U->H | accuracy: 74.72 -> 77.22 | Improved |
| use temporal discriminator instead of spatial discriminator (ablation) | TemRelation · UCF-HMDB full U->H | accuracy: 74.44 -> 74.72 | Improved |
| use temporal discriminator instead of spatial discriminator (ablation) | TemPooling · UCF-HMDB full U->H | accuracy: 71.11 -> 71.11 | No clear effect |
| use TemRelation instead of TemPooling with DANN | DANN · UCF-HMDB full U->H | accuracy: 71.11 -> 75.28 | Improved |
| use TemRelation instead of TemPooling with DANN | DANN · UCF-HMDB full H->U | accuracy: 75.13 -> 76.36 | Improved |
| apply TA3N instead of source-only training | TemRelation · Kinetics-Gameplay | accuracy: 17.22 -> 27.50 | Improved |
Closest papers by meaning
- Partial Video Domain Adaptation With Partial Adversarial Temporal Attentive Network (ICCV 2021 · similarity 0.87)
- Source-Free Video Domain Adaptation by Learning Temporal Consistency for Action Recognition (ECCV 2022 · similarity 0.87)
- Unsupervised Video Domain Adaptation for Action Recognition: A Disentanglement Perspective (NeurIPS 2023 · similarity 0.86)
- Temporal Alignment Networks for Long-Term Video (CVPR 2022 · similarity 0.86)
- Contrast and Mix: Temporal Contrastive Video Domain Adaptation with Background Mixing (NeurIPS 2021 · similarity 0.85)
- VidLA: Video-Language Alignment at Scale (CVPR 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.
