Research map / Label-efficient and robust learning / Domain adaptation and generalization
Contrast and Mix: Temporal Contrastive Video Domain Adaptation with Background Mixing
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 background mixing to temporal contrastive learning (ablation) | CoMix · UCF-HMDB average | accuracy: 85.8 -> 88.7 | Helped |
| add target pseudo-label supervised contrastive loss (ablation) | CoMix · UCF-HMDB average | accuracy: 85.8 -> 87.7 | Helped |
| add both BGM and TPL to TCL (ablation) | CoMix · Jester | accuracy: 57.5 -> 64.7 | Helped |
| background mixing instead of Video MixUp (ablation) | CoMix · UCF-HMDB average | accuracy: 88.4 -> 90.3 | Helped |
| background mixing instead of Video CutMix (ablation) | CoMix · Jester | accuracy: 58.6 -> 64.7 | Helped |
| add pseudo-labels (extra positives) for background-mixed videos (ablation) | CoMix · UCF-HMDB average | accuracy: 89.0 -> 90.3 | Helped |
| add contrastive loss on source domain (ablation) | CoMix · UCF-HMDB average | accuracy: 88.4 -> 90.3 | Helped |
| use random speed for slow branch (ablation) | CoMix · UCF-HMDB average | accuracy: 89.6 -> 90.3 | Helped |
| supervised contrastive on pseudo-labels instead of self-training cross-entropy (ablation) | CoMix · UCF-HMDB average | accuracy: 88.7 -> 90.3 | Helped |
| add GCN graph encoder (vs none) (ablation) | CoMix · UCF-HMDB average | accuracy: 88.1 -> 90.3 | Helped |
| GCN instead of LSTM temporal encoder (ablation) | CoMix · UCF-HMDB average | accuracy: 84.3 -> 90.3 | Helped |
| temporal median filtering instead of GMM for backgrounds (ablation) | CoMix · UCF-HMDB average | accuracy: 88.0 -> 90.3 | Helped |
Closest papers by meaning
- Discovering Informative and Robust Positives for Video Domain Adaptation (ICLR 2023 · similarity 0.92)
- Connect, Not Collapse: Explaining Contrastive Learning for Unsupervised Domain Adaptation (ICML 2022 · similarity 0.90)
- SimProF: A Simple Probabilistic Framework for Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.87)
- Link-based Contrastive Learning for One-Shot Unsupervised Domain Adaptation (CVPR 2025 · similarity 0.86)
- On Unsupervised Domain Adaptation: Pseudo Label Guided Mixup for Adversarial Prompt Tuning (AAAI 2024 · similarity 0.85)
- Unsupervised Video Domain Adaptation with Masked Pre-Training and Collaborative Self-Training (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.