Research map / Video / Video-language understanding
SVD: A Large-Scale Short Video Dataset for Near-Duplicate Video Retrieval
Qing-Yuan Jiang, Yi He, Gen Li, Jian Lin, Lei Li, Wu-Jun Li
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 |
|---|---|---|---|
| apply rotation transformation to positive videos | DML · SVD-Rotation | MAP (%): 78.47 -> 15.59 | Hurt |
| apply cropping transformation to positive videos | DML · SVD-Cropping | MAP (%): 78.47 -> 54.07 | Hurt |
| apply black border insertion to positive videos | DML · SVD-BlackBorder | MAP (%): 78.47 -> 68.17 | Hurt |
| apply speeding transformation to positive videos | DML · SVD-Speeding | MAP (%): 78.47 -> 76.70 | Hurt |
| apply rotation transformation to positive videos | HDML hashing 32 bits · SVD-Rotation | MAP (%): 31.59 -> 1.95 | Hurt |
| apply rotation transformation to positive videos | CNNL · SVD-Rotation | MAP (%): 55.55 -> 0.15 | Hurt |
| increase hash code length from 16 to 32 bits | LSH · SVD | MAP (%): 4.34 -> 28.36 | Improved |
| increase hash code length from 16 to 32 bits | HDML · SVD | MAP (%): 6.47 -> 31.59 | Improved |
| add Euclidean reranking of the top-N hashing results | HDML hashing 32 bits · SVD | MAP vs Top-100 MAP (%): 31.59 -> 78.63 | Improved |
| use hashing with reranking instead of real-value features | DML · SVD | retrieval time (ms): 2203.5 -> 174.8 | Improved |
| use deep metric learning instead of unsupervised aggregation | DML vs CNNL · SVD | MAP (%): 55.55 -> 78.47 | Improved |
| use hashing instead of real-value features (no reranking) | HDML 32 bits vs DML · SVD | MAP (%): 78.47 -> 31.59 | Hurt |
Closest papers by meaning
- MUVR: A Multi-Modal Untrimmed Video Retrieval Benchmark with Multi-Level Visual Correspondence (NeurIPS 2025 · similarity 0.84)
- EgoCVR: An Egocentric Benchmark for Fine-Grained Composed Video Retrieval (ECCV 2024 · similarity 0.84)
- Vid-SME: Membership Inference Attacks against Large Video Understanding Models (NeurIPS 2025 · similarity 0.83)
- VVS: Video-to-Video Retrieval with Irrelevant Frame Suppression (AAAI 2024 · similarity 0.82)
- VarCMP: Adapting Cross-Modal Pre-Training Models for Video Anomaly Retrieval (AAAI 2025 · similarity 0.82)
- Open-Vocabulary Video Anomaly Detection (CVPR 2024 · similarity 0.82)
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.
