lizhong323/LumiView

# [LumiView: A Synthetic Object Centric Dataset with Multiple View and Lighting](xxxx)
[Zhong Li](https://sites.google.com/site/lizhong19900216)¹,
[Liangchen Song](https://lsongx.github.io/)²,
[Zhang Chen](https://zhangchen8.github.io/)¹,
Xiangyu Du¹,
[Lele Chen](https://lelechen63.github.io/)¹,
[Junsong Yuan](https://cse.buffalo.edu/~jsyuan/)²,
[Yi Xu](https://www.linkedin.com/in/yi-xu-42654823/)¹,
¹OPPO US Research Center, ²University at Buffalo

<p align="center">
<img src='imgs/dataset_teaser.png' width="750"/>
</p>

# Introduction

> We used Blender’s physically based path tracer renderer and rendered 3 textured objects: `synthetic face`, `wood train`, and `face mask`. We set up 5 × 5 camera views on the front hemisphere, set 105 directional light sources around the full sphere, and render at a resolution of 800 × 800 pixels. Each camera differs by 10 degrees and each light source differs by 25 degrees on the sphere. Please download the dataset from this link: [BlenderData](xxxxx), and put them in the folder `data/BlenderData/`.

# Dataset structure
Each object has its own folder, and each folder contains the following files:
- `cam_data_label.npz`: camera parameters labeled in label['imgname'],label['intrinsic'] ,label['R_bcam'],w2c_T = label['T_bcam'],l_dirs = label['light_dir'], please refer to the github code[link](xxxxx) for more details.

- `xx_xx_xx_xx.png`: view and light index images.
```
root/
|-- FaceBase/
| |-- xx_xx_xx_xx.png # view and light index
| |-- cam_data_label.npz
|-- facecover/
|-- toytrain/
```

# Citation

```
@inproceedings{li2023relitneulf,
title={Relit-NeuLF: Efficient Novel View Synthesis with Neural 4D Light Field},
author={Li, Zhong, Song, Liangchen, Chen, Zhang, Du, Xiangyu, Chen, Lele, Yuan, Junsong, Xu, Yi},
booktitle={Proceedings of the 31th ACM International Conference on Multimedia},
year={2023}
}
```

原始信息汇总

LumiView: A Synthetic Object Centric Dataset with Multiple View and Lighting

简介

该数据集使用Blender的基于物理的路径追踪渲染器，渲染了三个纹理对象：合成面部、木制火车和面部口罩。在正面半球上设置了5×5的相机视角，并在全球周围设置了105个方向性光源，渲染分辨率为800×800像素。每个相机视角相差10度，每个光源相差25度。

数据集结构

每个对象有自己的文件夹，每个文件夹包含以下文件：

• cam_data_label.npz：包含相机参数，标签在label[imgname]、label[intrinsic]、label[R_bcam]、w2c_T = label[T_bcam]、l_dirs = label[light_dir]中。更多细节请参考GitHub代码。
• xx_xx_xx_xx.png：视角和光源索引图像。

目录结构如下：

root/ |-- FaceBase/ | |-- xx_xx_xx_xx.png # 视角和光源索引 | |-- cam_data_label.npz |-- facecover/ |-- toytrain/

引用

@inproceedings{li2023relitneulf, title={Relit-NeuLF: Efficient Novel View Synthesis with Neural 4D Light Field}, author={Li, Zhong, Song, Liangchen, Chen, Zhang, Du, Xiangyu, Chen, Lele, Yuan, Junsong, Xu, Yi}, booktitle={Proceedings of the 31th ACM International Conference on Multimedia}, year={2023} }