Research map / Video / Video-language understanding
MovieCuts: A New Dataset and Benchmark for Cut Type Recognition
Alejandro Pardo, Fabian Caba, Juan León Alcázar, Ali Thabet, Bernard Ghanem
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 |
|---|---|---|---|
| fine-tune the backbones on MovieCuts instead of training a linear classifier on frozen features (ablation) | Audio-Visual baseline · MovieCuts val | mAP: 30.82% to 46.57% | Improved |
| add the audio stream to the visual-only model (ablation) | fine-tuned visual model · MovieCuts val | mAP: 43.98% -> 46.57% | Improved |
| add audio to visual model for visually driven cuts (Cross Cut) (ablation) | fine-tuned AV model · MovieCuts val, Cross Cut | AP: 33.2 -> 31.1 | Hurt |
| add audio to visual model for audio-driven cuts (L Cut) (ablation) | fine-tuned AV model · MovieCuts val, L Cut | AP: 33.7 -> 43.3 | Improved |
| add Gradient Blending loss weights instead of equal weights (ablation) | fine-tuned AV model · MovieCuts val | mAP: 46.57% to 47.43% | Improved |
| scale the Gradient Blending weights by a factor of 3 (ablation) | AV + GB · MovieCuts val | mAP: 47.43% to 47.91% | Improved |
| use fixed window sampling around the cut instead of uniform sampling (ablation) | AV + Scaled GB · MovieCuts val | mAP: 47.17% -> 47.91% | Improved |
| use fixed window sampling instead of Gaussian sampling (ablation) | AV + Scaled GB · MovieCuts val | mAP: 47.44% -> 47.91% | Improved |
| replace BCE with Distribution-Balanced loss (ablation) | AV Scaled (no GB weights) · MovieCuts val | mAP: 47.4 -> 47.8 | Improved |
| add Distribution-Balanced loss to the scaled GB model (ablation) | AV + Scaled GB · MovieCuts val | mAP: 47.9 -> 47.9 | No clear effect |
| use the MovieCuts AV model to choose cut points instead of biased random cutting | automatic video editing · EditStock editing | F1 vs human edit: 77 -> 81 | Improved |
Closest papers by meaning
- VEU-Bench: Towards Comprehensive Understanding of Video Editing (CVPR 2025 · similarity 0.86)
- CineTechBench: A Benchmark for Cinematographic Technique Understanding and Generation (NeurIPS 2025 · similarity 0.85)
- Señorita-2M: A High-Quality Instruction-based Dataset for General Video Editing by Video Specialists (NeurIPS 2025 · similarity 0.84)
- MovieBench: A Hierarchical Movie Level Dataset for Long Video Generation (CVPR 2025 · similarity 0.84)
- VE-Bench: Subjective-Aligned Benchmark Suite for Text-Driven Video Editing Quality Assessment (AAAI 2025 · similarity 0.83)
- ShotBench: Expert-Level Cinematic Understanding in Vision-Language Models (NeurIPS 2025 · similarity 0.83)
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.
