Research map / Image generation / Image restoration
Learning Deep Local Features With Multiple Dynamic Attentions for Large-Scale Image Retrieval
Hui Wu, Min Wang, Wengang Zhou, Houqiang Li
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 dynamic attention (DA) head instead of 1x1 conv attention (ablation) | MDA, N=8 · ROxf Hard | mAP: 60.6 -> 62.2 | Improved |
| add dynamic attention (DA) head instead of l2 norm attention (ablation) | MDA, N=8 · ROxf Hard | mAP: 60.5 -> 62.2 | Improved |
| use DA instead of 1x1 conv + l2 norm attention (ablation) | MDA, N=8 · RPar Hard | mAP: 64.8 -> 66.2 | Improved |
| add stop gradient from attention branch to backbone (ablation) | MDA with DA, N=8 · ROxf Hard | mAP: 60.3 -> 62.2 | Improved |
| add stop gradient from attention branch to backbone (ablation) | MDA with DA, N=8 · RPar Hard | mAP: 64.0 -> 66.2 | Improved |
| add channel mapping layer to diversity regularization (ablation) | MDA, N=8 · ROxf Hard | mAP: 59.4 -> 62.2 | Improved |
| add channel mapping layer to diversity regularization (ablation) | MDA, N=8 · RPar Hard | mAP: 61.2 -> 66.2 | Improved |
| add diversity regularization to channel mapping (ablation) | MDA, N=8 · ROxf Hard | mAP: 60.2 -> 62.2 | Improved |
| add diversity regularization to channel mapping (ablation) | MDA, N=8 · RPar Hard | mAP: 64.7 -> 66.2 | Improved |
| increase head number from 1 to 8 (ablation) | MDA · ROxf-RPar Hard and Medium | mAP: 57.8 -> mAP rises until N=8 | Improved |
| increase head number to 10 or more (ablation) | MDA · ROxf-RPar | mAP | Hurt |
| single head DA instead of HOW attention | MDA N=1 vs HOW · Hard, ROxf | mAP: +0.9 mAP (57.8 vs HOW) | Improved |
Closest papers by meaning
- Deep Image Retrieval: Learning Global Representations for Image Search (ECCV 2016 · similarity 0.90)
- Learning Super-Features for Image Retrieval (ICLR 2022 · similarity 0.89)
- Large-Scale Image Retrieval With Attentive Deep Local Features (ICCV 2017 · similarity 0.89)
- Learning Token-Based Representation for Image Retrieval (AAAI 2022 · similarity 0.89)
- On the Connection between Local Attention and Dynamic Depth-wise Convolution (ICLR 2022 · similarity 0.88)
- End-to-End Learning of Deep Visual Representations for Image Retrieval (IJCV 2017 · similarity 0.88)
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.
