Research map / Detection and segmentation / Open-set and few-shot detection
Detect-To-Retrieve: Efficient Regional Aggregation for Image Search
Marvin Teichmann, Andre Araujo, Menglong Zhu, Jack Sim
Check 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 detector-selected regions and aggregate them with R-ASMK* | DELF-ASMK* · R-Oxf Hard | mAP: 41.0 -> 45.6 | Improved |
| Add detector-selected regions and aggregate them with R-ASMK* | DELF-ASMK* · R-Par Hard | mAP: 54.6 -> 57.7 | Improved |
| Retrain DELF on Google Landmarks | DELF-D2R-R-ASMK* · R-Oxf Hard | mAP: 45.6 -> 47.6 | Improved |
| Add spatial verification | DELF-GLD · R-Oxf Hard | mAP: 47.6 -> 52.4 | Improved |
| Add spatial verification | DELF-GLD · R-Par Hard | mAP: 61.3 -> 58.6 | Hurt |
| Add detector-selected regions for separate regional indexing (ablation) | DELF-ASMK* · R-Oxf Hard | mAP: +2.31% | Improved |
| Use average pooling instead of max pooling for regional search (ablation) | D2R-ASMK* · R-Oxf Hard | mAP: 40.35% | Mixed |
| Replace standard VLAD with R-VLAD regional aggregation (ablation) | DELF-VLAD · R-Oxf Hard | mAP: 30.17% -> 30.91%; 16.46% | Mixed |
| Replace R-ASMK with Naive-R-ASMK (ablation) | DELF-ASMK · R-Oxf Hard | mAP: 39.72% -> 31.42%; 9.2% | Hurt |
| Use R-ASMK* rather than Naive-R-ASMK with many selected regions (ablation) | DELF-ASMK* · R-Oxf Hard | mAP: 41.6% | Improved |
| Use spatial verification on R-Par (ablation) | DELF-GLD · R-Par Medium | mAP: 80.7 -> 80.2 | Hurt |
Closest papers by meaning
- On Train-Test Class Overlap and Detection for Image Retrieval (CVPR 2024 · similarity 0.89)
- Google Landmarks Dataset v2 - A Large-Scale Benchmark for Instance-Level Recognition and Retrieval (CVPR 2020 · similarity 0.89)
- Large-Scale Image Retrieval With Attentive Deep Local Features (ICCV 2017 · similarity 0.88)
- FORB: A Flat Object Retrieval Benchmark for Universal Image Embedding (NeurIPS 2023 · similarity 0.88)
- Instance-Level Composed Image Retrieval (NeurIPS 2025 · similarity 0.86)
- Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition (AAAI 2025 · similarity 0.86)
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.
