DLCA-Recon: Dynamic Loose Clothing Avatar Reconstruction from Monocular Videos

Authors: Chunjie Luo, Fei Luo, Yusen Wang, Enxu Zhao, Chunxia Xiao

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on public and our own datasets validate that our method can produce superior results for humans with loose clothing compared to the SOTA methods.
Researcher Affiliation Academia School of Computer Science, Wuhan University, Wuhan, China luochunjie@whu.edu.cn, luofei@whu.edu.cn, wangyusen@whu.edu.cn, zhaoenxu@whu.edu.cn, cxxiao@whu.edu.cn
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks (clearly labeled algorithm sections or code-like formatted procedures).
Open Source Code No The paper does not include an unambiguous statement that the authors are releasing the code for the methodology described, nor does it provide a direct link to a code repository.
Open Datasets Yes We evaluate our method on the Deep Cap dataset (Habermann et al. 2020), Dyna Cap dataset (Habermann et al. 2021) and our own captured data (LCJ, LYZ and ZJ).
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning.
Hardware Specification Yes The optimization takes 200 epochs (about 48 hours) on a single NVIDIA RTX 3090 GPU.
Software Dependencies No The paper mentions several software components like Py Torch3D, Py MAF, RVM, and SPIN, but does not provide specific version numbers for any of them or other underlying software dependencies.
Experiment Setup Yes The optimization takes 200 epochs (about 48 hours) on a single NVIDIA RTX 3090 GPU. Loss Implicit = losscolor + losseik + λlossnorm, where λ = 0.1. The optimizations for pose and skinning weights are disabled at the beginning of training, and they are gradually enabled when the non-rigid deformation network acquires a certain level of representation capacity.