Research map / Label-efficient and robust learning / Self-supervised and contrastive
Switchable Representation Learning Framework With Self-Compatibility
Shengsen Wu, Yan Bai, Yihang Lou, Xiongkun Linghu, Jianzhong He, Ling-Yu Duan
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 |
|---|---|---|---|
| Use SFSC instead of the unified model | ResNet-18 visual retrieval · Market1501, M(phi_1/16, phi) | mAP: 55.25 -> 69.43 | Improved |
| Use SFSC instead of the unified model | ResNet-18 visual retrieval · MSMT17, M(phi_1/16, phi) | mAP: 11.69 -> 28.16 | Improved |
| Use SFSC instead of the unified model | ResNet-18 visual retrieval · VeRi-776, M(phi_1/16, phi) | mAP: 44.79 -> 55.04 | Improved |
| Add uncertainty-based compatible loss (ablation) | SFSC · Market1501, M(phi_1/16, phi) | mAP: 66.41 -> 69.43 | Improved |
| Add uncertainty-based compatible loss (ablation) | SFSC · Market1501, M(phi, phi) | mAP: 77.87 -> 81.43 | Improved |
| Use gradient projection instead of summation aggregation (ablation) | SFSC · Market1501, M(phi, phi) | mAP: 75.38 -> 81.43 | Improved |
| Use gradient projection instead of summation aggregation (ablation) | SFSC · Market1501, M(phi_1/16, phi) | mAP: 68.08 -> 69.43 | Improved |
| Increase the number of crop ratios in W (ablation) | SFSC · Market1501, varied sub-model sizes | mAP | Mixed |
| Decrease the minimum crop ratio in W (ablation) | SFSC · Market1501, varied sub-model sizes | mAP | Mixed |
Closest papers by meaning
- Learning Compatible Embeddings (ICCV 2021 · similarity 0.85)
- CoReS: Compatible Representations via Stationarity (TPAMI 2023 · similarity 0.84)
- Seed the Views: Hierarchical Semantic Alignment for Contrastive Representation Learning (TPAMI 2022 · similarity 0.84)
- Stationary Representations: Optimally Approximating Compatibility and Implications for Improved Model Replacements (CVPR 2024 · similarity 0.84)
- Learning Compatible Multi-Prize Subnetworks for Asymmetric Retrieval (CVPR 2025 · similarity 0.83)
- In-Context Symmetries: Self-Supervised Learning through Contextual World Models (NeurIPS 2024 · similarity 0.83)
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.
