Research map / Video / Video-language understanding
Dual-Stream Knowledge-Preserving Hashing for Unsupervised Video Retrieval
Pandeng Li, Hongtao Xie, Jiannan Ge, Lei Zhang, Shaobo Min, Yongdong Zhang
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 |
|---|---|---|---|
| add dual-stream structure (ablation) | CNN encoder · FCVID 16 bits | MAP@5: 0.229 -> 0.273 | Improved |
| add dual-stream structure (ablation) | LSTM encoder · FCVID 16 bits | MAP@5: 0.227 -> 0.272 | Improved |
| add dual-stream structure (ablation) | Transformer encoder · FCVID 16 bits | MAP@5: 0.235 -> 0.297 | Improved |
| add dual-stream structure (ablation) | Transformer encoder · FCVID 64 bits | MAP@5: 0.477 -> 0.494 | Improved |
| add Gaussian-adaptive similarity graph instead of plain adjacency (ablation) | DKPH · FCVID 16 bits | MAP@5: 0.202 -> 0.297 | Improved |
| use Gaussian-adaptive graph instead of multiple matrices (ablation) | DKPH · FCVID 16 bits | MAP@5: 0.269 -> 0.297 | Improved |
| use dual-stream instead of twin bottlenecks (ablation) | DKPH · FCVID 16 bits | MAP@5: 0.228 -> 0.297 | Improved |
| add teacher-student distillation (ablation) | DKPH · FCVID 16 bits | MAP@5: 0.158 -> 0.297 | Improved |
| add binary structure similarity loss (ablation) | DKPH · FCVID 16 bits | MAP@5: 0.174 -> 0.297 | Improved |
| add visual embedding similarity loss (ablation) | DKPH · FCVID 16 bits | MAP@5: 0.209 -> 0.297 | Improved |
| add reconstruction loss to the student stream (ablation) | DKPH · FCVID 16 bits | MAP@5: 0.276 -> 0.297 | Improved |
| train on FCVID, test on YFCC vs same-dataset | DKPH vs BTH · YFCC cross-dataset 64 bits | MAP@20: -2.8% vs -5.7% | Improved |
Closest papers by meaning
- Neighborhood Preserving Hashing for Scalable Video Retrieval (ICCV 2019 · similarity 0.90)
- Efficient Self-Supervised Video Hashing with Selective State Spaces (AAAI 2025 · similarity 0.88)
- AutoSSVH: Exploring Automated Frame Sampling for Efficient Self-Supervised Video Hashing (CVPR 2025 · similarity 0.88)
- Deep Unsupervised Hashing via External Guidance (ICML 2025 · similarity 0.86)
- Mitigating Semantic Collapse in Partially Relevant Video Retrieval (NeurIPS 2025 · similarity 0.86)
- Effective Comparative Prototype Hashing for Unsupervised Domain Adaptation (AAAI 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.
