Research map / Video / Action recognition
A Local-to-Global Approach to Multi-Modal Movie Scene Segmentation
Anyi Rao, Linning Xu, Yu Xiong, Guodong Xu, Qingqiu Huang, Bolei Zhou, Dahua Lin
Check 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 multiple semantic elements to the place-feature base model (ablation) | LGSS · MovieScenes Test | AP: 19.5 -> 24.3 | Improved |
| Add BNet shot-boundary modeling (ablation) | LGSS · MovieScenes Test | AP: 24.3 -> 42.2 | Improved |
| Add the local sequence model (ablation) | LGSS · MovieScenes Test | AP: 42.2 -> 44.9 | Improved |
| Add global optimal grouping (ablation) | LGSS with local sequence model · MovieScenes Test | AP: 44.9 -> 47.1 | Improved |
| Add audio to place-only representations (ablation) | LGSS · MovieScenes Test | AP: 39.0 -> 43.4 | Improved |
| Add action to place-only representations (ablation) | LGSS · MovieScenes Test | AP: 39.0 -> 45.5 | Improved |
| Add cast to place-only representations (ablation) | LGSS · MovieScenes Test | AP: 39.0 -> 43.0 | Improved |
| Combine place, cast, action, and audio (ablation) | LGSS · MovieScenes Test | AP: 39.0 -> 47.1 | Improved |
| Use a longer temporal sequence window (ablation) | LGSS with BNet window 2 · MovieScenes Test | AP: 43.4 -> 46.5 | Improved |
| Use sequence length 20 instead of 10 (ablation) | LGSS with BNet window 4 · MovieScenes Test | AP: 47.1 -> 46.9 | Hurt |
| Increase global grouping iterations from 2 to 5 with 600 initial super shots (ablation) | LGSS global grouping · MovieScenes Test | AP: 46.3 -> 47.1 | Improved |
| Pretrain on MovieScenes (ablation) | LGSS · OVSD | AP: 76.2 -> 85.7 | Improved |
Closest papers by meaning
- Video Scene Segmentation with Genre and Duration Signals (ICLR 2026 · similarity 0.90)
- Towards Global Video Scene Segmentation with Context-Aware Transformer (AAAI 2023 · similarity 0.88)
- MEGA: Multimodal Alignment Aggregation and Distillation For Cinematic Video Segmentation (ICCV 2023 · similarity 0.88)
- MovieNet: A Holistic Dataset for Movie Understanding (ECCV 2020 · similarity 0.88)
- MovieBench: A Hierarchical Movie Level Dataset for Long Video Generation (CVPR 2025 · similarity 0.87)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 2024 · similarity 0.87)
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.
