README.md

metadata

license: cc-by-nc-4.0
tags:
  - LLIE
  - low-light
  - denoising
  - real-world

[WACV'26] Low-light Smartphone Dataset (LSD)

This is the official dataset proposed in our paper titled "Illuminating Darkness: Learning to Enhance Low-light Images In-the-Wild"

📄 Paper: arXiv
💻 Code: GitHub - LSD-TFFormer

Overview

We introduce LSD, the largest in-the-wild Single-Shot Low-Light Image Enhancement (SLLIE) dataset to date.

Dataset Structure

This repository contains the following training data files:

• patch_DLL_gtPatch.tar.gz
• patch_DLL_inputPatch.tar.gz
• patch_NLL_gtPatch.tar.gz
• patch_NLL_inputPatch.tar.gz

Categories

• DLL (Denoised Low-Light): For low-light enhancement training
• NLL (Noisy Low-Light): For joint denoising + enhancement training

File Organization

• inputPatch: Low-light input images
• gtPatch: Ground truth (well-lit) reference images

Usage

Extract the archives to access the training patches:

tar -xzf patch_DLL_gtPatch.tar.gz
tar -xzf patch_DLL_inputPatch.tar.gz
tar -xzf patch_NLL_gtPatch.tar.gz
tar -xzf patch_NLL_inputPatch.tar.gz

Dataset Status

✅ Training Dataset: Available (current files)
🔄 Test Dataset: Coming soon

Citation

You can cite our preprint as:

@article{sharif2025illuminating,
  title={Illuminating darkness: Enhancing real-world low-light scenes with smartphone images},
  author={Sharif, SMA and Rehman, Abdur and Abidin, Zain Ul and Naqvi, Rizwan Ali and Dharejo, Fayaz Ali and Timofte, Radu},
  journal={arXiv preprint arXiv:2503.06898},
  year={2025}
}