Research map / Video / Video segmentation and tracking
PAC-Net: Highlight Your Video via History Preference Modeling
Hang Wang, Penghao Zhou, Chong Zhou, Zhao Zhang, Xing Sun
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 DBC (user-level adaptive classifier) to the generic baseline (ablation) | PAC-NetG · PHD-GIFs (Dv3) | mAP: 15.28 -> 15.89 | Improved |
| add Mi-Hi mechanism with a shared classifier to the baseline (ablation) | PAC-NetG · PHD-GIFs (Dv3) | mAP: 15.28 -> 16.23 | Improved |
| use fine-grained Mi-Hi samples instead of concatenating averaged history (ablation) | PAC-NetAgg · PHD-GIFs (Dv3) | mAP: 15.67 -> 16.23 | Improved |
| concatenate averaged user history feature to the input feature (ablation) | PAC-NetG · PHD-GIFs (Dv3) | mAP: 15.28 -> 15.67 | Improved |
| combine DBC with Mi-Hi (full model) (ablation) | PAC-NetG · PHD-GIFs (Dv3) | mAP: 15.28 -> 17.51 | Improved |
| add non-highlighted history as DBC input (ablation) | PAC-Net · PHD-GIFs (Dv3) | mAP: 15.66 -> 17.51 | Improved |
| allow more user history videos (0 to all M) (ablation) | PAC-Net · PHD-GIFs (Dv3) | mAP: 15.28 -> 17.51 | Improved |
| increase allowed history videos from 3 to 5 (ablation) | PAC-Net · PHD-GIFs (Dv3) | mAP: 16.38 -> 17.02 | Improved |
| use PAC-Net instead of Adaptive-H-FCSN | personalized highlight detector · PHD-GIFs (Dv3) | mAP: 16.04 -> 17.51 | Improved |
| use PAC-Net with 5.89M parameters instead of Adaptive-H-FCSN with 197.36M | personalized highlight detector · PHD-GIFs (Dv3) | parameters (M): 197.36 -> 5.89 | Improved |
| use Adaptive-H-FCSN instead of FCSN on the harder Dv3 data | FCSN · PHD-GIFs (Dv3) | mAP: 15.09 -> 16.04 | Improved |
| transfer generic PAC-NetG trained on PHD-GIFs to SumMe without fine-tuning | PAC-NetG · SumMe | F-score: 44.4 -> 44.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Video2GIF — Video2GIF: Automatic Generation of Animated GIFs From Video (CVPR 2016) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PR-Net: Preference Reasoning for Personalized Video Highlight Detection (ICCV 2021 · similarity 0.92)
- Adaptive Video Highlight Detection by Learning from User History (ECCV 2020 · similarity 0.90)
- Aha! - Predicting What Matters Next: Online Highlight Detection Without Looking Ahead (NeurIPS 2025 · similarity 0.83)
- Exploring Domain Incremental Video Highlights Detection with the LiveFood Benchmark (AAAI 2024 · similarity 0.82)
- Bridging the Gap: A Unified Video Comprehension Framework for Moment Retrieval and Highlight Detection (CVPR 2024 · similarity 0.82)
- Watch Video, Catch Keyword: Context-aware Keyword Attention for Moment Retrieval and Highlight Detection (AAAI 2025 · 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.
