Research map / Vision-language and multimodal / Human and embodied modeling
End-to-End Learning of Deep Visual Representations for Image Retrieval
Albert Gordo, Jon Almazán, Jérôme Revaud, Diane Larlus
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 |
|---|---|---|---|
| finetune with ranking loss after classification warm-up (vs ImageNet weights) (ablation) | R-MAC ResNet101 · Oxford 5k | mAP: 69.4 -> 84.1 | Improved |
| use ranking loss instead of classification loss (ablation) | R-MAC VGG16 · Oxford 5k | mAP: 74.2 -> 79.9 | Improved |
| warm up with classification finetuning before ranking (ablation) | R-MAC VGG16 · Oxford 5k | mAP: 76.3 -> 79.9 | Improved |
| warm up with classification finetuning before ranking (ablation) | R-MAC ResNet101 · Oxford 5k | mAP: 83.4 -> 84.1 | Improved |
| train ranking on cleaned landmarks instead of full (ablation) | R-MAC · Oxford/Paris/Holidays | mAP | Improved |
| finetune with classification loss on clean instead of full landmarks (ablation) | R-MAC ResNet101 · Oxford 5k | mAP: 77.7 -> 78.5 | No clear effect |
| finetune for retrieval (ablation) | R-MAC ResNet101 · UKB | recall@4: 3.89 -> 3.83 | Hurt |
| replace rigid grid by 256 region proposals (ablation) | R-MAC VGG16 ImageNet · Oxford 5k | mAP: 60.3 -> 65.4 | Improved |
| replace rigid grid by 256 region proposals (ablation) | R-MAC ResNet101 finetuned · Oxford 5k | mAP: 84.1 -> 85.2 | Improved |
| replace rigid grid by 20 region proposals (ablation) | R-MAC ResNet101 finetuned · Holidays | mAP: 94.0 -> 92.0 | Hurt |
| add multi-resolution on query side (ablation) | Ft ResNet101 · Oxford 5k | mAP: 84.1 -> 84.9 | Improved |
| add multi-resolution on database side (ablation) | Ft ResNet101 · Oxford 5k | mAP: 84.1 -> 85.2 | Improved |
Closest papers by meaning
- Deep Image Retrieval: Learning Global Representations for Image Search (ECCV 2016 · similarity 0.95)
- CNN Image Retrieval Learns from BoW: Unsupervised Fine-Tuning with Hard Examples (ECCV 2016 · similarity 0.91)
- SIFT Meets CNN: A Decade Survey of Instance Retrieval (TPAMI 2017 · similarity 0.90)
- Fine-Tuning CNN Image Retrieval with No Human Annotation (TPAMI 2018 · similarity 0.90)
- Learning Super-Features for Image Retrieval (ICLR 2022 · similarity 0.90)
- Exploiting Distribution Constraints for Scalable and Efficient Image Retrieval (ICLR 2025 · 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.
