Pavilion-of-Prince-Teng
收藏资源简介:
TW-Pavilion数据集是论文《DANTE-W: Diffuse Albedo Neural Texturing in the Wild》(ECCV 2026)的配套数据集,旨在支持野外漫反射反照率神经纹理绘制的研究。该数据集基于GigaNVS和GigaLit数据集,提供了场景“滕王阁”的一个预览子集,完整数据集预计未来公开。数据内容包含多模态信息:多视角RGB图像(存储在images_4文件夹中)、相机参数(遵循MVSNet约定,存储在cams_4文件夹中)、漫反射反照率预测(存储在albedo_4文件夹中),以及对齐相机坐标的重建网格模型(以.obj和.mtl文件格式提供)。数据集规模较小(小于1K样本),适用于图像到三维任务,如神经渲染、纹理映射和三维重建。数据组织为分层结构,便于与官方DANTE-W代码库集成使用。
The TW-Pavilion dataset is a companion dataset to the paper DANTE-W: Diffuse Albedo Neural Texturing in the Wild (ECCV 2026), aimed at supporting research on diffuse albedo neural texturing in the wild. It is based on the GigaNVS and GigaLit datasets and provides a preview subset of the Tengwang Pavilion scene, with the full dataset expected to be released in the future. The data includes multimodal information: multi-view RGB images (stored in the images_4 folder), camera parameters (following the MVSNet convention, stored in the cams_4 folder), diffuse albedo predictions (stored in the albedo_4 folder), and reconstructed mesh models aligned with camera coordinates (provided in .obj and .mtl file formats). The dataset is small in scale (less than 1K samples) and suitable for image-to-3D tasks such as neural rendering, texture mapping, and 3D reconstruction. The data is organized in a hierarchical structure for easy integration with the official DANTE-W codebase.
数据集概述
- 名称:TW-Pavilion(The Pavilion of Prince Teng)
- 语言:英文
- 许可证:Apache-2.0
- 规模:样本数量小于1K
- 任务类别:图像到3D(image-to-3d)
- 标签:神经渲染、纹理映射、3D重建、多视图立体
来源与背景
- 该数据集是论文 DANTE-W: Diffuse Albedo Neural Texturing in the Wild(ECCV 2026)的一部分。
- 数据来源于 GigaNVS 和 GigaLit 数据集中的场景“The Pavilion of Prince Teng”的预览子集。
- 完整数据集将在后续公开发布。
数据内容与结构
数据集需按以下结构组织为文件夹 Pavilion_of_Prince_Teng:
data_noon/:主数据目录images_4/:存储采集的多视图RGB图像(.JPG格式)cams_4/:存储相机参数文件(.txt格式,遵循MVSNet约定)albedo_4/:存储漫反射反照率预测结果(.jpg格式)
.obj文件:场景的重建网格模型(与相机对齐),包括1_tex.obj、6.obj.mtl文件:材质定义文件,包括1_tex.mtl、6.mtl
引用
若使用该数据集,请引用以下论文: bibtex @InProceedings{Wang_2024_CVPR, author = {Wang, Guangyu and Zhang, Jinzhi and Wang, Fan and Huang, Ruqi and Fang, Lu}, title = {XScale-NVS: Cross-Scale Novel View Synthesis with Hash Featurized Manifold}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, month = {June}, year = {2024}, pages = {21029-21039} }




