Boosting Point Clouds Rendering via Radiance Mapping

Authors: Xiaoyang Huang, Yi Zhang, Bingbing Ni, Teng Li, Kai Chen, Wenjun Zhang

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We obtain a PSNR of 31.74 on Ne RF-Synthetic, 25.88 on Scan Net and 30.81 on DTU. Code and data are publicly available in https://github.com/seanywang0408/Radiance Mapping. ... Experiments ... Benchmark Evaluation ... Ablation Study
Researcher Affiliation Collaboration Xiaoyang Huang1*, Yi Zhang1*, Bingbing Ni1 , Teng Li2, Kai Chen3, Wenjun Zhang1 1Shanghai Jiao Tong University, Shanghai 200240, China 2Anhui University 3Shanghai AI Lab
Pseudocode No The paper describes the method and pipeline in text and uses figures, but it does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes Code and data are publicly available in https://github.com/seanywang0408/Radiance Mapping.
Open Datasets Yes We experiment on three datasets: Ne RF-Synthetic (Mildenhall et al.), Scan Net (Dai et al.) and DTU (Aanæs et al.).
Dataset Splits No The paper mentions 'training and testing views are fixed' and uses standard datasets like Ne RF-Synthetic, Scan Net, and DTU, but it does not explicitly provide specific percentages, counts, or a detailed methodology for creating train/validation/test splits within the paper's text.
Hardware Specification Yes All results are measured on a Ti TAN Xp with an image size of 800 800.
Software Dependencies No The paper mentions software components such as Open GL, Py Torch3D, and Adam optimizer, but does not provide specific version numbers for any of them.
Experiment Setup Yes We set the radius threshold τ as 5e-3 for Ne RF-Synthetic, 1.5e-2 for Scan Net, and 3e-3 for DTU. ... We use the Adam optimizer for training, with a batch size of 1. The initial learning rates of MLP and U-Net are 5e-4 and 1.5e-4 respectively, which are multiplied by 0.9999 in each step.