Research map / Image generation / Autoregressive and latent generation
Learning Super-Features for 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 FIRe Super-features instead of HOW local features | Image retrieval with R50− · R-Oxford | mAP: +3.5 med; +5.4 hard | Helped |
| Use FIRe Super-features instead of HOW local features | Image retrieval with R50− · R-Oxford + R1M | mAP: +2.9 med; +3.3 hard | Helped |
| Use FIRe Super-features instead of HOW local features | Image retrieval with R50− · R-Paris | mAP: +5.2 med; +9.9 hard | Helped |
| Use FIRe Super-features instead of HOW local features | Image retrieval with R50− · R-Paris + R1M | mAP: +9.2 med; +12.2 hard | Helped |
| Add attention decorrelation loss (ablation) | FIRe training · Landmark retrieval | mAP | Helped |
| Add direct contrastive loss on Super-features (ablation) | FIRe training · Landmark retrieval | mAP | Helped |
| Remove global loss while retaining Super-feature loss and attention decorrelation loss (ablation) | FIRe training · Landmark retrieval | mAP | Helped |
| Add global loss to Super-feature and attention decorrelation losses (ablation) | FIRe training · Landmark retrieval | mAP | No clear effect |
| Add reciprocal-nearest-neighbor filtering to matching pairs (ablation) | FIRe pair selection · Landmark retrieval | mAP: SfM-120k val: 68.3 -> 79.6 | Helped |
| Add same-ID constraint along with reciprocal matching (ablation) | FIRe pair selection · Landmark retrieval | mAP: SfM-120k val: 79.6 -> 89.9 | Helped |
| Replace residual template update with a GRU (ablation) | LIT (T=3) · R-Oxford | mAP: 71.3 -> 66.7 med; 43.9 -> 41.4 hard | Hurt |
| Replace residual template update with a GRU (ablation) | LIT (T=3) · R-Paris | mAP: 78.7 -> 77.2 med; 59.4 -> 57.8 hard | Hurt |
Closest papers by meaning
- Learning Token-Based Representation for Image Retrieval (AAAI 2022 · similarity 0.87)
- Exploiting Distribution Constraints for Scalable and Efficient Image Retrieval (ICLR 2025 · similarity 0.86)
- Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition (AAAI 2025 · similarity 0.85)
- Instance-Level Composed Image Retrieval (NeurIPS 2025 · similarity 0.85)
- On Train-Test Class Overlap and Detection for Image Retrieval (CVPR 2024 · similarity 0.85)
- QuARI: Query Adaptive Retrieval Improvement (NeurIPS 2025 · 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.