Research map / Label-efficient and robust learning / Domain adaptation and generalization
Granularity-Aware Adaptation for Image Retrieval over Multiple Tasks
Jon Almazán, Byungsoo Ko, Geonmo Gu, Diane Larlus, Yannis Kalantidis
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 pseudo-granularity adaptors with attention-consistency fusion | DINO ViT-S · Flowers | mAP: 57.0 -> 61.9 | Improved |
| add pseudo-granularity adaptors with attention-consistency fusion | DINO ViT-S · Products | mAP: 31.5 -> 33.6 | Improved |
| pretrain adaptors on pseudo-labels instead of random initialization (ablation) | GrAPPa fusion · CUB | mAP: 4.5 -> 24.1 | Improved |
| pretrain adaptors on pseudo-labels instead of random initialization (ablation) | GrAPPa fusion · Cars | mAP: 1.7 -> 4.0 | Improved |
| use neighbor attention consistency instead of transformation consistency for fusion (ablation) | GrAPPa · CUB | mAP: 23.5 -> 24.1 | Improved |
| use neighbor attention consistency instead of transformation consistency for fusion (ablation) | GrAPPa · Products | mAP: 33.2 -> 33.6 | Improved |
| use learned attention fusion instead of average pooling (ablation) | GrAPPa · Products | mAP: 34.4 -> 33.6 | Hurt |
| use learned attention fusion instead of average pooling (ablation) | GrAPPa · CUB | mAP: 23.1 -> 24.1 | Improved |
| fuse all eight adaptor sets rather than pick best single set (ablation) | GrAPPa-AC · Food-101 | mAP: 18.4 -> 19.3 | Improved |
| use finest pseudo-granularity (P8) instead of P3 only (ablation) | single adaptor set · CUB | mAP: 24.5 -> 6.6 | Hurt |
| use finest pseudo-granularity (P8) instead of P1 (ablation) | single adaptor set · Products | mAP: 29.6 -> 43.7 | Improved |
| add a single randomly initialized adaptor (ablation) | DINO · CUB | mAP: 22.9 -> 7.8 | Hurt |
Closest papers by meaning
- Coarse-to-Fine: Learning Compact Discriminative Representation for Single-Stage Image Retrieval (ICCV 2023 · similarity 0.88)
- Unicom: Universal and Compact Representation Learning for Image Retrieval (ICLR 2023 · similarity 0.86)
- Fine-Grained Image Retrieval via Dual-Vision Adaptation (AAAI 2026 · similarity 0.86)
- NUDGE: Lightweight Non-Parametric Fine-Tuning of Embeddings for Retrieval (ICLR 2025 · similarity 0.85)
- Towards Efficient and Effective Text-to-Video Retrieval with Coarse-to-Fine Visual Representation Learning (AAAI 2024 · similarity 0.85)
- MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions (ICML 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.
