Voxurf: Voxel-based Efficient and Accurate Neural Surface Reconstruction
Authors: Tong Wu, Jiaqi Wang, Xingang Pan, Xudong XU, Christian Theobalt, Ziwei Liu, Dahua Lin
ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show that Voxurf achieves high efficiency and high quality at the same time. On the DTU benchmark, Voxurf achieves higher reconstruction quality with a 20x training speedup compared to previous fully implicit methods. We conduct experiments on the DTU (Jensen et al., 2014) and Blended MVS (Yao et al., 2020) datasets for quantitative and qualitative evaluations. |
| Researcher Affiliation | Collaboration | Tong Wu1,2, Jiaqi Wang1B, Xingang Pan3, Xudong Xu2, Christian Theobalt3, Ziwei Liu4, Dahua Lin1,2,5B 1Shanghai AI Laboratory, 2The Chinese University of Hong Kong, 3Max Planck Institute for Informatics, 4S-Lab, Nanyang Technological University, 5Centre of Perceptual and Interactive Intelligence {wt020, xx018, dhlin}@ie.cuhk.edu.hk, wangjiaqi@pjlab.org.cn, {xpan,theobalt}@mpi-inf.mpg.de, ziwei.liu@ntu.edu.sg |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our code is available at https://github.com/wutong16/Voxurf. |
| Open Datasets | Yes | We conduct experiments on the DTU (Jensen et al., 2014) and Blended MVS (Yao et al., 2020) datasets for quantitative and qualitative evaluations. |
| Dataset Splits | No | The paper mentions using DTU and Blended MVS datasets but does not provide specific details about the training, validation, or test splits (e.g., percentages or sample counts) in the main text. It defers to supplementary materials for 'further descriptions of the datasets, baseline methods, and implementation details.' |
| Hardware Specification | Yes | All the training times are tested on a single Nvidia A100 GPU. ...reducing the training time from over 5 hours to 15 minutes on a single Nvidia A100 GPU. |
| Software Dependencies | No | The paper describes the software components of its model (e.g., MLPs, voxel grids) but does not specify any software dependencies with version numbers (e.g., Python, PyTorch, CUDA versions). |
| Experiment Setup | No | The paper describes the components of its loss function and mentions loss weights (λ0, λtv, λs) but does not provide specific numerical values for these weights or other common hyperparameters like learning rate, batch size, or number of epochs in the main text. It refers to supplementary materials for 'implementation details'. |