Research map / Video / Video-language understanding
Neighborhood Preserving Hashing for Scalable Video Retrieval
Shuyan Li, Zhixiang Chen, Jiwen Lu, Xiu Li, Jie Zhou
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 neighborhood information reconstruction loss (ablation) | NPH · FCVID (32 bits) | mAP@20: 0.240 -> 0.246 | Improved |
| add neighborhood information reconstruction loss (ablation) | NPH · FCVID (64 bits) | mAP@20: 0.286 -> 0.294 | Improved |
| use neighborhood attention instead of anchor concatenation (ablation) | NPH · FCVID (32 bits) | mAP@20: 0.214 -> 0.240 | Improved |
| use neighborhood attention instead of anchor concatenation (ablation) | NPH · FCVID (64 bits) | mAP@20: 0.254 -> 0.286 | Improved |
| replace self-attention with neighborhood attention (visual loss only) (ablation) | NPH · FCVID (32 bits) | mAP@20: 0.187 -> 0.201 | Improved |
| replace self-attention with neighborhood attention (visual loss only) (ablation) | NPH · FCVID (64 bits) | mAP@20: 0.239 -> 0.238 | No clear effect |
| add neighborhood similarity loss to self-attention encoder (ablation) | NPH · FCVID (32 bits) | mAP@20: 0.187 -> 0.210 | Improved |
| add neighborhood similarity loss to self-attention encoder (ablation) | NPH · FCVID (64 bits) | mAP@100: 0.113 -> 0.162 | Improved |
| add neighborhood attention and reconstruction losses on top of SelfAtt+Ls (ablation) | NPH · FCVID (32 bits) | mAP@20: 0.210 -> 0.240 | Improved |
| use fully convolutional encoder instead of LSTM (ablation) | FullCNN · FCVID (32 bits) | mAP@20: 0.189 -> 0.175 | No clear effect |
| NPH versus SSVH on unseen classes | NPH · FCVID unseen classes | mAP@5: 0.300 -> 0.306 | Improved |
| NPH versus SSVH after training on train25 | NPH · FCVID train25 | mAP@20: 0.185 -> 0.193 | Improved |
Closest papers by meaning
- Dual-Stream Knowledge-Preserving Hashing for Unsupervised Video Retrieval (ECCV 2022 · similarity 0.90)
- Contrastive Masked Autoencoders for Self-Supervised Video Hashing (AAAI 2023 · similarity 0.89)
- Efficient Self-Supervised Video Hashing with Selective State Spaces (AAAI 2025 · similarity 0.89)
- Deep Hashing Network for Efficient Similarity Retrieval (AAAI 2016 · similarity 0.89)
- AutoSSVH: Exploring Automated Frame Sampling for Efficient Self-Supervised Video Hashing (CVPR 2025 · similarity 0.88)
- Hashing with Binary Matrix Pursuit (ECCV 2018 · similarity 0.88)
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.
