Research map / Video / Video segmentation and tracking
Mr. HiSum: A Large-scale Dataset for Video Highlight Detection and Summarization
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 on Mr. HiSum then fine-tune on target | SL-module · TVSum average | MAP: 71.3±1.5 -> 73.0±0.7 | Helped |
| train on Mr. HiSum then fine-tune on target | SL-module · YouTube highlight average | MAP: 59.9±1.7 -> 62.0±1.3 | Helped |
| zero-shot transfer from Mr. HiSum without fine-tuning | SL-module · TVSum average | MAP: 71.3±1.5 -> 69.5±0.4 | Hurt |
| zero-shot transfer from Mr. HiSum without fine-tuning | SL-module · YouTube highlight dog | MAP: 48.9±3.7 -> 56.1±2.1 | Helped |
| fine-tune on small target after Mr. HiSum pretraining | SL-module · YouTube highlight skating | MAP: 43.5±1.3 -> 42.0±2.6 | Hurt |
| train on Mr. HiSum instead of TVSum train set | PGL-SUM · TVSum | F1: 55.5±0.7 -> 57.1±0.7 | Helped |
| train on Mr. HiSum instead of TVSum train set | VASNet · TVSum | F1: 54.2±0.9 -> 57.1±1.0 | Helped |
| train on Mr. HiSum instead of SumMe train set | PGL-SUM · SumMe | F1: 41.7±3.2 -> 42.3±2.1 | Helped |
| train on Mr. HiSum instead of SumMe train set | VASNet · SumMe | F1: 41.9±3.3 -> 42.6±1.3 | Helped |
| zero-shot from Mr. HiSum on categories with few Mr. HiSum videos | SL-module · YouTube highlight surfing | MAP: 74.7±2.4 -> 66.9±1.6 | Hurt |
| increase validation set size from small to 1,000 or more (ablation) | PGL-SUM · Mr. HiSum validation | F1 stability | Helped |
| increase training set size for Food category (ablation) | PGL-SUM · Mr. HiSum Food | F1 | Helped |
Closest papers by meaning
- Bridging the Gap: A Unified Video Comprehension Framework for Moment Retrieval and Highlight Detection (CVPR 2024 · similarity 0.88)
- Watch Video, Catch Keyword: Context-aware Keyword Attention for Moment Retrieval and Highlight Detection (AAAI 2025 · similarity 0.87)
- Exploring Domain Incremental Video Highlights Detection with the LiveFood Benchmark (AAAI 2024 · similarity 0.87)
- ShotVL: Human-Centric Highlight Frame Retrieval via Language Queries (AAAI 2025 · similarity 0.86)
- TR-DETR: Task-Reciprocal Transformer for Joint Moment Retrieval and Highlight Detection (AAAI 2024 · similarity 0.86)
- MMSum: A Dataset for Multimodal Summarization and Thumbnail Generation of Videos (CVPR 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.