Research map / Detection and segmentation / CNN features
Large-Scale Image Retrieval With Attentive Deep Local Features
Hyeonwoo Noh, Andre Araujo, Jack Sim, Tobias Weyand, Bohyung Han
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 |
|---|---|---|---|
| add attention-based keypoint selection (ablation) | DELF without fine-tuning · Google-Landmarks (large-scale retrieval) | precision-recall | Improved |
| add landmark fine-tuning of descriptors (ablation) | DELF · Google-Landmarks (large-scale retrieval) | precision-recall | Improved |
| use attention instead of fine-tuning alone (ablation) | DELF · Google-Landmarks (large-scale retrieval) | precision-recall | Improved |
| add query expansion | DIR · Google-Landmarks (large-scale retrieval) | precision-recall | Hurt |
| add query expansion | DIR · Oxford5k | mAP (%): 86.1 -> 87.1 | Improved |
| add query expansion | siaMAC · Paris106k | mAP (%): 76.3 -> 79.8 | Improved |
| late fusion of DELF with DIR+QE | DIR+QE · Oxford5k | mAP (%): 87.1 -> 90.0 | Improved |
| late fusion of DELF with DIR+QE | DIR+QE · Paris106k | mAP (%): 91.8 -> 92.8 | Improved |
| use DELF alone instead of DIR | global descriptor · Paris6k | mAP (%): 94.5 -> 85.0 | Hurt |
| train attention jointly with descriptors instead of two-step (ablation) | DELF · Google-Landmarks | retrieval quality | Hurt |
| add random image rescaling during attention training (ablation) | attention model · Google-Landmarks | retrieval quality | Improved |
Closest papers by meaning
- Unifying Deep Local and Global Features for Image Search (ECCV 2020 · similarity 0.91)
- Deep Image Retrieval: Learning Global Representations for Image Search (ECCV 2016 · similarity 0.90)
- Learning Deep Local Features With Multiple Dynamic Attentions for Large-Scale Image Retrieval (ICCV 2021 · similarity 0.89)
- Detect-To-Retrieve: Efficient Regional Aggregation for Image Search (CVPR 2019 · similarity 0.88)
- Learning Super-Features for Image Retrieval (ICLR 2022 · similarity 0.88)
- On Train-Test Class Overlap and Detection for Image Retrieval (CVPR 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.
