OpsiClear/NGS
收藏Hugging Face2025-10-17 更新2025-11-01 收录
下载链接：
https://hf-mirror.com/datasets/OpsiClear/NGS
下载链接
链接失效反馈官方服务：
资源简介：
Noise Guided Splatting (NGS)透明度数据集是一个用于研究和基准测试3D高斯散点绘制中透明度错误问题的数据集。它由四个子集组成：stones和objects是原始高分辨率数据集，分别包含超过100个石质样本和多种日常物品；DTU和OmniObject3D是从流行基准测试中增强的子集，包含了用于评估透明度的噪声高斯填充文件。数据集旨在提供具有明显透明度问题的数据，并使用表面透明度评分(SOS)指标进行量化。
The Noise Guided Splatting (NGS) Transparency Datasets are designed for researching and benchmarking the false transparency artifact in 3D Gaussian Splatting reconstructions. It consists of four subsets: the original high-resolution datasets stones and objects, each containing over 100 stone specimens and a variety of everyday items, respectively; and the augmented subsets DTU and OmniObject3D from popular benchmarks, which include noise Gaussian infills for evaluating transparency. The datasets aim to provide data with pronounced transparency issues and are quantified using the proposed Surface Opacity Score (SOS) metric.
提供机构：
OpsiClear
