MultiAct: Long-Term 3D Human Motion Generation from Multiple Action Labels
Authors: Taeryung Lee, Gyeongsik Moon, Kyoung Mu Lee
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our Multi Act outperforms the best combination of previous SOTA methods to generate long-term motion from multiple action labels, besides handling such problem within a single model. The experimental comparison is conducted on the quality of action motion and transition in single-step and long-term generations. |
| Researcher Affiliation | Collaboration | 1 IPAI, Seoul National University, Korea 2 Dept. of ECE & ASRI, Seoul National University, Korea 3 Meta Reality Labs Research |
| Pseudocode | No | The paper includes diagrams illustrating the model (e.g., Figure 2 and Figure 3), but it does not contain a dedicated pseudocode or algorithm block. |
| Open Source Code | Yes | Code is publicly available in https://github.com/TaeryungLee/MultiActRELEASE. |
| Open Datasets | Yes | BABEL (Punnakkal et al. 2021) is the only dataset that consists of a long-term human motion with sequential action labels. |
| Dataset Splits | No | The paper states 'We use training and validation split for training and testing of our model, respectively.' and 'The detail of data sampling is illustrated in supplementary materials.' However, it does not provide specific percentages or sample counts for the splits in the main text. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments. |
| Software Dependencies | No | The paper mentions several architectural components and models (e.g., Transformer, CVAE, SMPL-H) but does not list any specific software libraries or their version numbers used in the implementation or experimentation. |
| Experiment Setup | No | The paper describes the model architecture and loss functions but does not provide specific hyperparameter values (e.g., learning rate, batch size, number of epochs) or detailed training configurations in the main text. |