Research map / Vision-language and multimodal / Cross-modal learning
Deep Discriminative Structure Proxy Hashing for Cross-modal Retrieval
Kun Cheng, Qibing Qin, Lei Huang
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 |
|---|---|---|---|
| use DDSPH instead of DDWSH | CLIP hashing · MIRFLICKR-25K I2T 128 bits | mAP@all: 87.00 -> 87.53 | Improved |
| use DDSPH at 16 bits on MS COCO | DDSPH · MS COCO I2T 16 bits | mAP@all: 72.84 (DHaPH) vs 69.57 | Hurt |
| use all proxies (r=1.0) instead of sparse proxy selection (ablation) | DDSPH · MIRFLICKR-25K 16 bits | mAP@all | Hurt |
| increase number of proxies N and ratio r (ablation) | DDSPH · MIRFLICKR-25K 16 bits | mAP@all | No clear effect |
| add sample-proxy alignment loss La (ablation) | DDSPH · MIRFLICKR-25K | mAP@all | Improved |
| add structure and preservation loss terms (Lr, Lp) (ablation) | DDSPH · MIRFLICKR-25K | mAP@all | Improved |
| use hard positive mask instead of soft or no mask (ablation) | DDSPH · MIRFLICKR-25K | mAP@all | Improved |
| use dynamic loss normalization instead of fixed weights 1/1/1 (ablation) | DDSPH · MIRFLICKR-25K I2T 32 bits | mAP: 85.52 -> 85.99 | Improved |
| use dynamic loss normalization instead of fixed weights 1/1/1 (ablation) | DDSPH · MS COCO I2T 32 bits | mAP: 74.12 -> 74.59 | Improved |
| use DDSPH under label noise rate 0.8 | DDSPH vs DDBH · MIRFLICKR-25K I2T 16 bits | mAP: 76.50 -> 81.16 | Improved |
| sample-specific proxy structuring | DDSPH · training efficiency | training time | Improved |
Closest papers by meaning
- Mutual Quantization for Cross-Modal Search With Noisy Labels (CVPR 2022 · similarity 0.90)
- Dynamic Masking and Auxiliary Hash Learning for Enhanced Cross-Modal Retrieval (NeurIPS 2025 · similarity 0.89)
- PromptHash:Affinity-Prompted Collaborative Cross-Modal Learning for Adaptive Hashing Retrieval (CVPR 2025 · similarity 0.89)
- Deep Evidential Hashing for Trustworthy Cross-Modal Retrieval (AAAI 2025 · similarity 0.88)
- Lightweight Contrastive Distilled Hashing for Online Cross-modal Retrieval (AAAI 2025 · similarity 0.88)
- Dual Self-Paced Cross-Modal Hashing (AAAI 2024 · 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.
