“There are two ways of spreading light: to be the candle or the mirror that receives it.” ― Edith Wharton (1862-1937)

• Please check our Arxiv, videos (Bilibili).
• DarkGS code (sister repo): DarkGS DarkGS relighting preview 👇 (Novel view rendering sequence)

pip install git+https://github.com/princeton-vl/lietorch.git PySide6

1. Clone the repo:

https://github.com/tyz1030/neuralight.git

2. Run gui.py:

python3 gui.py

  Our toy dataset will be automatically loaded.
  3. Load our toy data checkpoint by simply click "LOAD" button on the gui.
  4. Check the (checkbox next to the) parameters you want to optimize and uncheck those to be freezed.
  5. Click "START" to start finetuning the model.
  6. Click "SAVE" to save your finetuned model.
  7. For more instructions see the following illustration:

An example calibration target can be downloaded here but feel free to generate your own target. We use moms-apriltag.

Arxiv

@misc{zhang2024darkgs,
      title={DarkGS: Learning Neural Illumination and 3D Gaussians Relighting for Robotic Exploration in the Dark}, 
      author={Tianyi Zhang and Kaining Huang and Weiming Zhi and Matthew Johnson-Roberson},
      year={2024},
      eprint={2403.10814},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

• This work is supported by NOAA.
• Copyright 2024 Kaining Huang and Tianyi Zhang, Carnegie Mellon University. All rights reserved.