Research map / Label-efficient and robust learning / Recognition and long-tail
Boundary-Aware Backward-Compatible Representation via Adversarial Learning in Image Retrieval
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 |
|---|---|---|---|
| Use AdvBCT instead of the strongest listed baseline | Backward-compatible retrieval training · GLDv2, Extended-class | P1-score: 51.51 -> 52.35 | Helped |
| Use AdvBCT instead of the strongest listed baseline | Backward-compatible retrieval training · GLDv2, Extended-data | P1-score: 53.52 -> 53.45 | Hurt |
| Use AdvBCT instead of the strongest listed baseline | Backward-compatible retrieval training · GLDv2, Extended-backbone (data) | P1-score: 53.25 -> 53.01 | Hurt |
| Use AdvBCT instead of the strongest listed baseline | Backward-compatible retrieval training · GLDv2, Extended-backbone (class) | P1-score: 52.15 -> 52.83 | Helped |
| Add elastic p2s loss (ablation) | Classification-plus-adversarial AdvBCT · RParis, Extended-data cross-test | mAP: 4.4 -> 78.55 | Helped |
| Add elastic p2s loss (ablation) | Classification-plus-adversarial AdvBCT · ROxford, Extended-data cross-test | mAP: 2.11 -> 52.31 | Helped |
| Add elastic p2s loss (ablation) | Classification-plus-adversarial AdvBCT · RParis, Extended-class cross-test | mAP: 5.67 -> 77.16 | Helped |
| Add elastic p2s loss (ablation) | Classification-plus-adversarial AdvBCT · ROxford, Extended-class cross-test | mAP: 2.34 -> 54.82 | Helped |
| Add elastic p2s loss (ablation) | Classification-plus-adversarial AdvBCT · ROxford, Extended-data self-test | mAP: 63.92 -> 62.13 | Hurt |
| Add adversarial loss (ablation) | Classification-plus-elastic-p2s AdvBCT · RParis, Extended-data | mAP: 82.12 -> 82.78; 77.18 -> 78.55 | Helped |
| Add adversarial loss (ablation) | Classification-plus-elastic-p2s AdvBCT · ROxford, Extended-data | mAP: 61.16 -> 62.13; 51.63 -> 52.31 | Mixed |
Closest papers by meaning
- Forward Compatible Training for Large-Scale Embedding Retrieval Systems (CVPR 2022 · similarity 0.90)
- Exploiting Distribution Constraints for Scalable and Efficient Image Retrieval (ICLR 2025 · similarity 0.87)
- NUDGE: Lightweight Non-Parametric Fine-Tuning of Embeddings for Retrieval (ICLR 2025 · similarity 0.85)
- Stationary Representations: Optimally Approximating Compatibility and Implications for Improved Model Replacements (CVPR 2024 · similarity 0.85)
- LORE: Lagrangian-Optimized Robust Embeddings for Visual Encoders (NeurIPS 2025 · similarity 0.85)
- Collapse-Aware Triplet Decoupling for Adversarially Robust Image Retrieval (ICML 2024 · similarity 0.84)
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.