Research map / Detection and segmentation / CNN features
DOLG: Single-Stage Image Retrieval With Deep Orthogonal Fusion of Local and Global Features
Min Yang, Dongliang He, Miao Fan, Baorong Shi, Xuetong Xue, Fu Li, Errui Ding, Jizhou Huang
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 the local branch (fused orthogonally with global feature) (ablation) | DOLG R50 · Roxf Medium | mAP: 78.21 -> 80.50 | Improved |
| add the local branch (ablation) | DOLG R50 · Rpar Medium | mAP: 89.00 -> 89.81 | Improved |
| add multi-atrous module (ablation) | DOLG R50 · Roxf Hard | mAP: 58.36 -> 58.82 | Improved |
| add multi-atrous module (ablation) | DOLG R50 · Roxf Easy | mAP: 93.48 -> 93.17 | Hurt |
| add self-attention block (ablation) | DOLG R50 · Roxf Hard | mAP: 55.34 -> 58.82 | Improved |
| fuse at res3 instead of res4 (ablation) | DOLG R50 · Roxf Hard | mAP: 58.13 -> 58.82 | Improved |
| fuse at res3 instead of res4 (ablation) | DOLG R50 · Rpar Hard | mAP: 77.92 -> 77.70 | Hurt |
| fuse both res3 and res4 instead of res3 only (ablation) | DOLG R50 · Roxf Hard | mAP: 58.82 -> 57.08 | No clear effect |
| use orthogonal fusion instead of concatenation (ablation) | DOLG R50 · Roxf Hard | mAP: 56.55 -> 58.82 | Improved |
| use orthogonal fusion instead of Hadamard product (ablation) | DOLG R50 · Roxf Hard | mAP: 56.76 -> 58.82 | Improved |
| use GeM global and average pooling in fusion instead of GeM/GeM (ablation) | DOLG R50 · Roxf Hard | mAP: 55.30 -> 58.82 | Improved |
| use GeM instead of average pooling in the global branch (average in fusion) (ablation) | DOLG R50 · Roxf Hard | mAP: 44.88 -> 58.82 | Improved |
Closest papers by meaning
- Unifying Deep Local and Global Features for Image Search (ECCV 2020 · similarity 0.90)
- Deep Image Retrieval: Learning Global Representations for Image Search (ECCV 2016 · similarity 0.89)
- End-to-End Learning of Deep Visual Representations for Image Retrieval (IJCV 2017 · similarity 0.88)
- Coarse-to-Fine: Learning Compact Discriminative Representation for Single-Stage Image Retrieval (ICCV 2023 · similarity 0.88)
- Learning Spatial-context-aware Global Visual Feature Representation for Instance Image Retrieval (ICCV 2023 · similarity 0.87)
- Instance-Level Composed Image Retrieval (NeurIPS 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.
