Research map / Architectures and efficiency / RNNs and normalization
Efficient Indexing of Billion-Scale Datasets of Deep Descriptors
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace IMI's orthogonal product decomposition with non-orthogonal codebooks | IMI indexing · DEEP1B | average distance to closest cell centroid: 0.321 -> 0.272; reduce average distance by 15% | Helped |
| Add learned cluster-wise linear coefficients to NO-IMI (ablation) | NO-IMI indexing · DEEP1B | average distance to closest cell centroid: 0.272 -> 0.255; improvement by 21% compared to IMI | Helped |
| Replace IMI's orthogonal product decomposition with non-orthogonal codebooks | IMI indexing · SIFT1B | average distance to closest cell centroid: 35923 -> 35207; reduce average distance by just three percent | Helped |
| Add learned cluster-wise linear coefficients to NO-IMI (ablation) | NO-IMI indexing · SIFT1B | average distance to closest cell centroid: 35207 -> 34981 | Helped |
| Use non-orthogonal indexing instead of IMI | billion-scale retrieval · DEEP1B | candidate-list length at recall 0.5: eight time more compact | Helped |
| Use non-orthogonal indexing instead of IMI | retrieval with reranking · DEEP1B | recall: up to six absolute percent; 13% of relative improvement | Helped |
| Use non-orthogonal indexing instead of IMI | billion-scale retrieval · SIFT1B | candidate-list quality | No clear effect |
Closest papers by meaning
- Learning a Complete Image Indexing Pipeline (CVPR 2018 · similarity 0.85)
- Efficient Large-Scale Similarity Search Using Matrix Factorization (CVPR 2016 · similarity 0.85)
- Local Descriptors Optimized for Average Precision (CVPR 2018 · similarity 0.84)
- Learning Super-Features for Image Retrieval (ICLR 2022 · similarity 0.84)
- Exploiting Distribution Constraints for Scalable and Efficient Image Retrieval (ICLR 2025 · similarity 0.84)
- Adaptive Retrieval and Scalable Indexing for k-NN Search with Cross-Encoders (ICLR 2024 · similarity 0.83)
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.