DiffusionLight-LoRA-Trainer: a code for training LoRA use in DIffusionLight

Project Page | Main Repository

We provide a code and training set for training LoRA using in DiffusionLight.

Table of contents

• TL;DR
• Dataset
• Installation

• Citation

Getting started

bash final_train_cont_largeT@900.sh

wait over a night on Single RTX 3090 until it create folder ThisIsTheFinal-lora-hdr-continuous-largeT@900/0_-5/checkpoint-2500 which we provide in main repository

Dataset

We use the Text2Light to generate HDR Environment map using the prompt which also generated by ChatGPT-3.5.

However, you can download a pre-generated dataset at HuggingFace or OneDrive (133 GB) which contain 1412 images pair of Low-FOV LDR and HDR Environment map

To download the dataset from Huggingface, simply use git clone (please make sure that you have git lfs enabled)

git clone https://huggingface.co/datasets/DiffusionLight/text2dataset

Installation

Please follow the instructions in our main repository's installation guide

Citation

@inproceedings{Phongthawee2023DiffusionLight,
    author = {Phongthawee, Pakkapon and Chinchuthakun, Worameth and Sinsunthithet, Nontaphat and Raj, Amit and Jampani, Varun and Khungurn, Pramook and Suwajanakorn, Supasorn},
    title = {DiffusionLight: Light Probes for Free by Painting a Chrome Ball},
    booktitle = {ArXiv},
    year = {2023},
}

Visit us 🦉