Research map / Label-efficient and robust learning / Domain adaptation and generalization
ELViS: Efficient Visual Similarity from Local Descriptors that Generalizes Across Domains
Pavel Suma, Giorgos Kordopatis-Zilos, Yannis Kalantidis, Giorgos Tolias
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 |
|---|---|---|---|
| remove dustbins from optimal transport (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP avg: 53.9 -> 30.4 | Improved |
| use descriptor-dependent dustbin gains instead of a scalar gain (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP avg: 51.5 -> 53.9 | Improved |
| add learnable vote function f (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP OOD: 53.8 -> 55.0 | Improved |
| add learnable vote function f (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP ID: 50.5 -> 50.8 | Hurt |
| add learnable loss function g during training (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP avg: 48.5 -> 53.9 | Improved |
| add f and g together (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP avg: 48.2 -> 53.9 | Improved |
| add learnable descriptor projection (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP avg: 50.8 -> 53.9 | Improved |
| use lambda 0.01 instead of 0.1 for OT entropy regularization (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP avg: 53.9 -> 47.4 | Hurt |
| use lambda 1.0 instead of 0.1 for OT entropy regularization (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP avg: 53.9 -> 48.4 | Hurt |
| make lambda learnable (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP avg: 53.9 -> 54.0 | No clear effect |
| increase Sinkhorn iterations from 1 to 20 (ablation) | ELViS · 8 datasets (GLDv2-trained) | mAP avg: 52.7 -> 54.1 | Improved |
| replace AMES with ELViS (similarity-space model) for re-ranking, trained on SOP | AMES · out-of-domain datasets (SOP-trained) | mAP OOD: 41.4 -> 51.0 | Improved |
Closest papers by meaning
- Efficient Discovery and Effective Evaluation of Visual Perceptual Similarity: A Benchmark and Beyond (ICCV 2023 · similarity 0.88)
- Beyond Instance-Level Image Retrieval: Leveraging Captions to Learn a Global Visual Representation for Semantic Retrieval (CVPR 2017 · similarity 0.87)
- Interpretable Measures of Conceptual Similarity by Complexity-Constrained Descriptive Auto-Encoding (CVPR 2024 · similarity 0.85)
- Instance-Level Composed Image Retrieval (NeurIPS 2025 · similarity 0.85)
- The Many Senses of Visual Similarity: A Text-Prompted Image Perceptual Metric (NeurIPS 2026 · similarity 0.85)
- LOCORE: Image Re-ranking with Long-Context Sequence Modeling (CVPR 2025 · similarity 0.85)
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.
