Research map / Label-efficient and robust learning / Recognition and long-tail
On Trivial Solution and High Correlation Problems in Deep Supervised Hashing
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 |
|---|---|---|---|
| train an ensemble of sub-models on different data subsets, each producing k/m bits, instead of one model | deep supervised hashing (triplet loss, AlexNet) · AwA, 128 bits | mean absolute correlation (mAC): 0.73 -> 0.25 | Helped |
| ensemble DSH instead of DNNH | deep supervised hashing · CIFAR10, 128 bits | mAP: 0.6531 -> 0.7051 | Helped |
| ensemble DSH instead of DNNH | deep supervised hashing · ImageNet, 128 bits | mAP: 0.5576 -> 0.6156 | Helped |
| ensemble DSH instead of DNNH | deep supervised hashing · AwA, 128 bits | mAP: 0.5607 -> 0.6078 | Helped |
| ensemble DSH instead of DNNH at short code length | deep supervised hashing · ImageNet, 32 bits | mAP: 0.4931 -> 0.4946 | No clear effect |
| increase hashcode length from 32 to 128 bits | Ensemble DSH vs other DSH · benchmarks | mAP gain: +8.07% (ours) vs 3.86%, 4.33%, 3.10% (DeSH, DNNH, DRSCH) | Helped |
| freeze conv1-conv3 layers shared across sub-models (ablation) | Ensemble DSH · ImageNet 96 bits | mAP: 0.5871 -> 0.5717 (pair-wise); 0.6022 -> 0.5831 (triplet) | No clear effect |
| freeze through conv5 or deeper layers (ablation) | Ensemble DSH · ImageNet 96 bits | mAP: 0.5871 -> 0.5173 (pair-wise, conv5); 0.6022 -> 0.4378 (triplet, fc7) | Hurt |
Closest papers by meaning
- Deep Supervised Hashing for Fast Image Retrieval (CVPR 2016 · similarity 0.92)
- Deep Supervised Discrete Hashing (NeurIPS 2017 · similarity 0.92)
- Deep Hashing Network for Efficient Similarity Retrieval (AAAI 2016 · similarity 0.90)
- One Loss for All: Deep Hashing with a Single Cosine Similarity based Learning Objective (NeurIPS 2021 · similarity 0.90)
- Pairwise-Label-Based Deep Incremental Hashing with Simultaneous Code Expansion (AAAI 2024 · similarity 0.88)
- Deep Graph Online Hashing for Multi-Label Image Retrieval (AAAI 2025 · similarity 0.87)
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.