Learning 3D Garment Animation from Trajectories of A Piece of Cloth

Authors: YIDI SHAO, Chen Change Loy, Bo Dai

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that EUNet effectively delivers the energy gradients due to the deformations.
Researcher Affiliation Collaboration Yidi Shao1 Chen Change Loy1 Bo Dai2,3 1S-Lab, Nanyang Technological University 2The University of Hong Kong, 3Shanghai Artificial Intelligence Laboratory
Pseudocode No The paper describes methods using text and mathematical equations but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes Code is available at https://github.com/ftbabi/EUNet_NeurIPS2024.git .
Open Datasets Yes To train our EUNet, we collect the dynamics of a piece of square cloth made of common materials, i.e., silk, leather, cotton, and denim. ... We evaluate the animations constrained by EUNet on ground truth garments from Cloth3D [2].
Dataset Splits Yes We generate 800 sequences of different clothes for training and 200 sequences for test, with a length of 30 frames for each sequence.
Hardware Specification Yes We train our model for six epochs on V100. ... All experiments are run on NVIDIA A100-SXM4-80GB.
Software Dependencies No The paper mentions software like Blender and deep learning components (e.g., MLP, Adam optimizer) but does not provide specific version numbers for these software or any key libraries.
Experiment Setup Yes We apply four blocks of multi-layer perceptron (MLP) with dimensions 128. Each block of MLP consists of 2 fully connected layers... We set λ = 106 to ensure sufficient regularization. We adopt the Adam optimizer with an initial learning rate of 0.0002, with a decreasing factor of 0.5 every four epochs. The batch size is set to 4. We train our model for six epochs on V100.