Research map / Label-efficient and robust learning / Recognition and long-tail
Contrastive Quantization with Code Memory for Unsupervised Image Retrieval
Read the paperCheck 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 |
|---|---|---|---|
| Use MeCoQ instead of CIBHash | unsupervised image retrieval · Flickr25K, CIFAR-10 (I), CIFAR-10 (II), NUS-WIDE | MAP: 3.52, 6.92, 2.24 and 1.46 | Helped |
| Add debiasing to correct sampling bias (ablation) | MeCoQ · Flickr25K, CIFAR-10 (I), CIFAR-10 (II), NUS-WIDE | MAP: 1.52; 2.66; 1.28; 1.48 | Helped |
| Add codeword diversity regularization (ablation) | MeCoQ · Flickr25K, CIFAR-10 (I), CIFAR-10 (II), NUS-WIDE | MAP: 24.18; 33.54; 28.42; 24.87 | Helped |
| Add soft quantization code memory (ablation) | MeCoQ · Flickr25K, CIFAR-10 (I), CIFAR-10 (II), NUS-WIDE | MAP: 2.21; 2.10; 2.87; 3.03 | Helped |
| Use soft code memory rather than feature memory (ablation) | MeCoQ · Flickr25K, CIFAR-10 (I), CIFAR-10 (II), NUS-WIDE | MAP: 2.10; 2.12; 0.77; 2.26 | Helped |
| Use soft code memory rather than hard code memory (ablation) | MeCoQ · Flickr25K, CIFAR-10 (I), CIFAR-10 (II), NUS-WIDE | MAP: 18.61; 29.09; 26.54; 9.93 | Helped |
| Delay memory-augmented training instead of using memory from the start (ablation) | MeCoQ · Flickr25K, CIFAR-10 (I), CIFAR-10 (II), NUS-WIDE | MAP: 3.27; 1.92; 2.67; 0.77 | Helped |
| Increase code-memory size from 384 to 896 slots (ablation) | MeCoQ · CIFAR-10 (I), 32 bits | MAP: 69.74 -> 69.17 | Hurt |
Closest papers by meaning
- Deep Quantization Network for Efficient Image Retrieval (AAAI 2016 · similarity 0.89)
- Deep Unsupervised Hashing via External Guidance (ICML 2025 · similarity 0.86)
- Pairwise-Label-Based Deep Incremental Hashing with Simultaneous Code Expansion (AAAI 2024 · similarity 0.86)
- HiHPQ: Hierarchical Hyperbolic Product Quantization for Unsupervised Image Retrieval (AAAI 2024 · similarity 0.85)
- Learning Attribute-Aware Hash Codes for Fine-Grained Image Retrieval via Query Optimization (ICML 2025 · similarity 0.85)
- An Asymmetric Augmented Self-Supervised Learning Method for Unsupervised Fine-Grained Image Hashing (CVPR 2024 · 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.