kewangtt / srlefnip

NightCampus dataset

Citation

@ARTICLE{9655477,
  author={Wang, Ke and Shen, Shaojie},
  journal={IEEE Robotics and Automation Letters}, 
  title={Semi-Supervised Learning: Structure, Reflectance and Lighting Estimation From a Night Image Pair}, 
  year={2022},
  volume={7},
  number={2},
  pages={976-983},
  doi={10.1109/LRA.2021.3136307}}

1. Introduction

The NightCampus dataset is a real night dataset with 12-bit raw images.

The dataset contains 220 scenes, and every scene contains two to four pairs of long-exposure images, over 20 pairs of short-exposure images and a dense point cloud. All short-exposure image pairs have an identical exposure time. The clean images for every scene are generated by averaging the sequence of the short-exposure images.

2. Usage

Demo code

We provide a Python demo.

python ./Tools/demo.py

Exposure Times & Intrinsic

Exposure times are listed in

./Tools/List.txt

Camera intrinsics are recorded in

./Tools/parameters.txt

3. Example

A left-hand image with a long exposure time

A left-hand clean image with a short exposure time

A left-hand depth image

A left-hand normal image

4. License

The open-source version is licensed under the GNU General Public License Version 3 (GPLv3).