Research map / Video / Video-language understanding
The Anatomy of Video Editing: A Dataset and Benchmark Suite for AI-Assisted Video Editing
Dawit Mureja Argaw, Fabian Caba, Joon-Young Lee, Markus Woodson, In So Kweon
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 logit adjustment for long-tail labels (single-task, video+audio) (ablation) | shot attribute classifier · AVE shot attributes classification (test) | average accuracy: 45.9 -> 57.3 | Improved |
| add logit adjustment (multi-task, video) (ablation) | shot attribute classifier · AVE shot attributes classification (test) | average accuracy: 44.4 -> 56.5 | Improved |
| add audio features to video input for sound source (ablation) | shot attribute classifier · AVE shot attributes classification (test), sound source | accuracy: 32.6 -> 38.9 | Improved |
| train each attribute classifier separately instead of multi-task (ablation) | shot attribute classifier · AVE shot attributes classification (test) | average accuracy: 56.7 -> 57.3 | Improved |
| use early fusion of shot inputs instead of late feature fusion (ablation) | shot ordering · AVE shot sequence ordering (test, audio+video) | accuracy: 24.4 -> 30.7 | Improved |
| add audio to video input (ablation) | Baseline-II shot ordering · AVE shot sequence ordering (test) | accuracy: 27.4 -> 30.7 | Improved |
| restrict to sequences containing an insert shot | Baseline-II shot ordering · AVE shot sequence ordering (test, audio+video) | accuracy: 30.7 -> 39.5 | Improved |
| restrict to intensify pattern sequences | Baseline-II shot ordering · AVE shot sequence ordering (test, audio+video) | accuracy: 30.7 -> 48.0 | Improved |
| use in-batch negatives in addition to in-sequence negatives (ablation) | next shot selection · AVE next shot selection (test, audio+video) | accuracy: 38.7 -> 41.4 | Improved |
| add attributes of neighboring shots as input (frame) (ablation) | missing shot attribute predictor · AVE missing shot attributes (test), shot size | accuracy: 32.4 -> 44.6 | Improved |
| add attributes of neighboring shots as input (video+audio) (ablation) | missing shot attribute predictor · AVE missing shot attributes (test), shot angle | accuracy: 35.8 -> 39.0 | Improved |
| pretrain ResNet-101 on shot attribute task instead of ImageNet (parameter-free OPTICS) | camera setup clustering · AVE camera setup clustering, OPTICS | NMI: 0.718 -> 0.802 | Improved |
Closest papers by meaning
- VEU-Bench: Towards Comprehensive Understanding of Video Editing (CVPR 2025 · similarity 0.89)
- EditBoard: Towards a Comprehensive Evaluation Benchmark for Text-Based Video Editing Models (AAAI 2025 · similarity 0.88)
- Learning Action and Reasoning-Centric Image Editing from Videos and Simulation (NeurIPS 2024 · similarity 0.87)
- VE-Bench: Subjective-Aligned Benchmark Suite for Text-Driven Video Editing Quality Assessment (AAAI 2025 · similarity 0.87)
- Señorita-2M: A High-Quality Instruction-based Dataset for General Video Editing by Video Specialists (NeurIPS 2025 · similarity 0.87)
- VINCIE-NExT: Unlocking Video Editing from Images via In-Context Modeling (NeurIPS 2026 · 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.
