Research map / Video / Video generation
LAMP: Language-Assisted Motion Planning for Controllable Video Generation
Muhammed Burak Kizil, Enes Sanli, Niloy J. Mitra, Erkut Erdem, Aykut Erdem, Duygu Ceylan
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 |
|---|---|---|---|
| predict DSL motion programs instead of binned per-frame trajectories (ablation) | LAMP · Procedural dataset test, free-form camera (DSL vs trajectory) | coarse translation F1: 0.847 -> 0.996 | Improved |
| predict DSL motion programs instead of binned per-frame trajectories (ablation) | LAMP · Procedural dataset test, free-form camera (DSL vs trajectory) | fine translation F1: 0.781 -> 0.966 | Improved |
| predict DSL motion programs instead of binned per-frame trajectories (ablation) | LAMP · Procedural dataset test, free-form camera (DSL vs trajectory) | coarse rotation F1: 0.730 -> 0.950 | Improved |
| predict DSL motion programs instead of binned per-frame trajectories (ablation) | LAMP · Procedural dataset test, free-form camera (DSL vs trajectory) | fine rotation F1: 0.626 -> 0.753 | Improved |
| predict DSL motion programs instead of binned per-frame trajectories (ablation) | LAMP · Procedural dataset test, free-form camera (DSL vs trajectory) | rotation MAE per axis (degrees): 8.333 -> 3.507 | Improved |
| fine-tune on DSL-converted DataDoP with regenerated captions | LAMP · DataDoP camera trajectories | revised F1: 0.763 -> 0.776 | Improved |
| use regenerated captions instead of original DataDoP captions in fine-tuning (ablation) | LAMP · DataDoP camera trajectories | revised F1: 0.613 -> 0.776 | Improved |
| train from scratch on DSL-converted DataDoP instead of fine-tuning (ablation) | LAMP · DataDoP camera trajectories | revised F1: 0.776 -> 0.776 | No clear effect |
| fine-tune on DSL-converted ET with regenerated captions | LAMP · ET dataset, mixed split | F1 score: 0.769 -> 0.779 | Improved |
| use regenerated captions instead of original ET captions in fine-tuning (ablation) | LAMP · ET dataset, pure split | F1 score: 0.666 -> 0.978 | Improved |
| train on DataDoP from scratch instead of procedural data only (ablation) | LAMP · ET dataset, pure split | F1 score: 0.976 -> 0.967 | No clear effect |
| replace GenDoP baseline with LAMP (no DataDoP training) | LAMP · DataDoP camera trajectories | revised F1: 0.360 -> 0.763 | Improved |
Closest papers by meaning
- LaMP: Language-Motion Pretraining for Motion Generation, Retrieval, and Captioning (ICLR 2025 · similarity 0.90)
- Auteur: Language-Driven Cinematographic Framing for Human-Centric Video Generation (NeurIPS 2026 · similarity 0.88)
- AvatarGPT: All-in-One Framework for Motion Understanding Planning Generation and Beyond (CVPR 2024 · similarity 0.88)
- Video Language Planning (ICLR 2024 · similarity 0.87)
- Generating Human Motion in 3D Scenes from Text Descriptions (CVPR 2024 · similarity 0.87)
- LAMP: Learn A Motion Pattern for Few-Shot Video Generation (CVPR 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.
