AMD: Anatomical Motion Diffusion with Interpretable Motion Decomposition and Fusion
Authors: Beibei Jing, Youjia Zhang, Zikai Song, Junqing Yu, Wei Yang
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on datasets with relatively more complex motions, such as CLCD1 and CLCD2, demonstrate that our AMD significantly outperforms existing state-of-the-art models. |
| Researcher Affiliation | Academia | Beibei Jing, Youjia Zhang, Zikai Song, Junqing Yu, Wei Yang * Huazhong University of Science and Technology, Wuhan, China {jingbeibei,youjiazhang, skyesong, yjqing, weiyangcs}@hust.edu.cn |
| Pseudocode | No | The paper does not contain any pseudocode or clearly labeled algorithm blocks. Figure 2 provides an architectural overview, not pseudocode. |
| Open Source Code | No | The paper does not include any statement or link indicating that the source code for the methodology is openly available. |
| Open Datasets | Yes | Human ML3D (Guo et al. 2022a) and KIT Motion-Language (Mandery et al. 2015) have become essential in the text-to-motion task. |
| Dataset Splits | No | The paper mentions using specific datasets for testing (SLCD1, SLCD2) and refers to general training, but does not explicitly provide specific percentages or sample counts for training, validation, or test splits. |
| Hardware Specification | No | The computation is completed in the HPC Platform of Huazhong University of Science and Technology. (This is a general statement about a computing platform but lacks specific hardware details like GPU/CPU models or memory.) |
| Software Dependencies | Yes | Specifically, we leverage the power of a Large Language Model, i.e., a fine-tuned Chat GPT-3.5 model, for decomposing complex action text descriptions... We use the Sentence-BERT model all-Mini LM-L6-v2 to compute the semantic embeddings of texts and anatomical scripts. |
| Experiment Setup | Yes | Specifically, for unconditional generation, we randomly mask 10% of the textual conditions and related action information. Additionally, we approximate the probability distribution x1:N 0 using a mixture of elements, and we random select pθ1 and pθ2... This fine-tuning enables the model to strike a balance between Diversity and Fidelity, allowing for a trade-off that enhances performance. |