Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses
Authors: Seungwoo Yoo, Juil Koo, Kyeongmin Yeo, Minhyuk Sung
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments with the Deform Things4D and Human datasets demonstrate state-of-the-art performance in pose transfer and the ability to generate diverse deformed shapes with various objects and poses. |
| Researcher Affiliation | Academia | Seungwoo Yoo Juil Koo Kyeongmin Yeo Minhyuk Sung KAIST {dreamy1534,63days,aaaaa,mhsung}@kaist.ac.kr |
| Pseudocode | No | The paper does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks, nor does it present structured steps in a code-like format. |
| Open Source Code | No | We will publicly release the code upon acceptance. |
| Open Datasets | Yes | For the animal shapes, we utilize animation sequences from the Deforming Things4D-Animals dataset [23]. For humanoids, we use SMPL [28, 31]... Additionally, we collect 9 stylized character meshes from the Mixamo dataset [1]. |
| Dataset Splits | No | The paper describes data used for training and testing, but it does not explicitly define a validation set or specific train/validation/test split percentages across the datasets used for general model training. |
| Hardware Specification | Yes | Our experiments are conducted on RTX 3090 GPUs (24 GB VRAM) and A6000 GPUs (48 GB VRAM). |
| Software Dependencies | No | The paper mentions using 'Point Transformer layers from Zhao et al.[56] and Tang et al.[44]' and that 'network architectures for our cascaded diffusion models... are adapted from Koo et al. [21]', but it does not provide specific version numbers for these or other software libraries or dependencies. |
| Experiment Setup | Yes | These models operate over T = 1000 timesteps with a linear noise schedule ranging from β1 = 1 10 4 to βT = 5 10 2. For model training, we employ the ADAM optimizer at a learning rate of 1 10 3 and standard parameters. For per-identity refinement modules, we set λlap = 1.0, λedge = 1.0, and λreg = 5 10 2 during training. |