Research map / 3D vision / 3D generation
Text Motion Translator: A Bi-Directional Model for Enhanced 3D Human Motion Generation from Open-Vocabulary Descriptions
Yijun Qian, Jack Urbanek, Alexander Hauptmann, Jungdam Won
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 LLM text augmentation with filtering to TMT (ablation) | TMT · BABEL unseen actions | APEr: 0.510 -> 0.327 | Improved |
| add LLM text augmentation with filtering to TMT (ablation) | TMT · BABEL full validation | APEr: 0.288 -> 0.237 APEr; 0.142 -> 0.144 AVEr | Mixed |
| use LLM augmentation without filtering heuristics (ablation) | TMT · BABEL full validation | APEr: 0.237 -> 0.349 | Hurt |
| add filtering post-processing to LLM-augmented descriptions (ablation) | TMT · BABEL seen actions | APEr: 0.309 -> 0.186 | Improved |
| add multi-task regularization (Text2Text, Motion2Motion, Motion2Text) over text2motion only (ablation) | TMT · BABEL unseen actions | APEr: 0.374 -> 0.327 | Improved |
| add multi-task regularization over text2motion only (ablation) | TMT · BABEL seen actions | APEr: 0.203 -> 0.186 | Improved |
| add LaViMo pretraining to motion VQ-VAE (ablation) | TMT · BABEL | APEr: 0.297 -> 0.237 | Improved |
| add LaViMo pretraining on top of AMASS (ablation) | TMT · BABEL unseen actions | APEr: 0.553 -> 0.327 | Improved |
| add LaViMo pretraining on top of AMASS (ablation) | TMT · BABEL seen actions | APEr: 0.153 -> 0.186 | Hurt |
| add LaViMo pretraining (ablation) | TMT · HumanML3D | FID: 0.310 -> 0.184 | Improved |
| use TMT instead of T2M-GPT on HumanML3D | TMT · HumanML3D | FID: 0.116 -> 0.184 | Hurt |
Closest papers by meaning
- Generating Diverse and Natural 3D Human Motions From Text (CVPR 2022 · similarity 0.92)
- The Language of Motion: Unifying Verbal and Non-verbal Language of 3D Human Motion (CVPR 2025 · similarity 0.91)
- SnapMoGen: Human Motion Generation from Expressive Texts (NeurIPS 2025 · similarity 0.90)
- Generating Human Motion in 3D Scenes from Text Descriptions (CVPR 2024 · similarity 0.90)
- ReMoGPT: Part-Level Retrieval-Augmented Motion-Language Models (AAAI 2025 · similarity 0.90)
- VimoRAG: Video-based Retrieval-augmented 3D Motion Generation for Motion Language Models (NeurIPS 2025 · similarity 0.89)
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.
