Research map / Label-efficient and robust learning / Recognition and long-tail
SIFT Meets CNN: A Decade Survey of Instance Retrieval
Liang Zheng, Yi Yang, Qi Tian
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 |
|---|---|---|---|
| rotate Holidays images to upright | retrieval methods · Holidays | mAP: 2-3% mAP improvement | Improved |
| replace pre-trained VGG16 with fine-tuned ResNet101 in Deep Retrieval | Deep Retrieval · Oxford5k | mAP: 83.1 -> 86.1 | Improved |
| add Hamming embedding with WGC to a large-codebook AKM baseline | BoW retrieval · Oxford5k | mAP: 49.3 -> 61.5 | Improved |
| add burstiness handling to HE+WGC | HE+WGC · Oxford5k | mAP: 61.5 -> 64.7 | Improved |
| add spatial verification/query expansion to Fine Vocabulary | Fine Voc. · Oxford5k | mAP: 74.2 -> 84.9 | Improved |
| fine-tune CNN on landmark data with MAC layer (SiaMAC) instead of pre-trained R-MAC | VGG16 pooled features · Oxford5k | mAP: 66.9 -> 77.0 | Improved |
| increase HE signature length from 64 to 128 bits | Hamming embedding · Oxford5k | accuracy | Improved |
| increase HE signature to 256 bits | Hamming embedding · Oxford5k | recall | Hurt |
| increase codebook size in large/medium SIFT codebooks | SIFT BoW · Oxford5k | mAP | Mixed |
| add HE to medium-sized codebooks | SIFT BoW · Oxford5k | mAP stability | Improved |
| reduce compact vector dimension below 256 or 128 | compact representations · Oxford5k/Holidays | mAP: 256, 128 | Hurt |
| fine-tune CNN on Landmarks dataset | AlexNet retrieval · Ukbench | mAP | Hurt |
Closest papers by meaning
- Deep Learning for Instance Retrieval: A Survey (TPAMI 2022 · similarity 0.93)
- Deep Image Retrieval: Learning Global Representations for Image Search (ECCV 2016 · similarity 0.91)
- End-to-End Learning of Deep Visual Representations for Image Retrieval (IJCV 2017 · similarity 0.90)
- Instance-Level Composed Image Retrieval (NeurIPS 2025 · similarity 0.89)
- ILIAS: Instance-Level Image retrieval At Scale (CVPR 2025 · similarity 0.86)
- Exploiting Distribution Constraints for Scalable and Efficient Image Retrieval (ICLR 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.
