Skip to main content

Datasets for Low-Light Enhancement

Created on March 31|Last edited on March 31

Table of Contents



LOL Dataset

The LOL dataset is composed of 500 low-light and normal-light image pairs and is divided into 485 training pairs and 15 testing pairs. The low-light images contain noise produced during the photo capture process. Most of the images are indoor scenes. All the images have a resolution of 400×600. The dataset was introduced in the paper Deep Retinex Decomposition for Low-Light Enhancement.

LoL
Artifact overview
Type
dataset
Created At
January 24th, 2023
Description
Versions
Version
Aliases
Logged By
Tags
Created
TTL Remaining
# of Consuming Runs
Size
m.Google-Drive
m.Original-Dataset
0
latest
v0
Tue Jan 24 2023
Inactive
130
348.1MB
https://drive.google.com/open?id=157bjO1_cFuSd0HWDUuAmcHRJDVyWpOxB
https://daooshee.github.io/BMVC2018website/
Loading...

Citation

@misc{
1808.04560,
Author = {Chen Wei and Wenjing Wang and Wenhan Yang and Jiaying Liu},
Title = {Deep Retinex Decomposition for Low-Light Enhancement},
Year = {2018},
Eprint = {arXiv:1808.04560},
}




Exclusively Dark (ExDark) Image Dataset

Research works on low-light enhancement have seen steady growth, particularly in the field of image enhancement, but there is still a lack of a go-to dataset as a benchmark. This dataset is aimed primarily to provide a dataset for benchmarking object detection under challenging low-light situations. This dataset has a collection of 7363 low-light images from very low-light environments to twilight (i.e 10 different conditions), and 12 object classes (as to PASCAL VOC) annotated on both image class level and local object bounding boxes.

Exclusively-Dark-Image-Dataset
Artifact overview
Type
dataset
Created At
March 29th, 2023
Description
Versions
Version
Aliases
Logged By
Tags
Created
TTL Remaining
# of Consuming Runs
Size
m.source
0
latest
v0
Wed Mar 29 2023
Inactive
11
1.5GB
CDN: https://drive.google.com/file/d/1BHmPgu8EsHoFDDkMGLVoXIlCth2dW6Yx Paper: http://cs-chan.com/doc/cviu.pdf GitHub: https://github.com/cs-chan/Exclusively-Dark-Image-Dataset/tree/master/Dataset
Loading...

Citation

@misc{
1805.11227,
Author = {Yuen Peng Loh and Chee Seng Chan},
Title = {Getting to Know Low-light Images with The Exclusively Dark Dataset},
Year = {2018},
Eprint = {arXiv:1805.11227},
}



EnlightenGAN Dataset

EnlightenGAN, as proposed by the paper EnlightenGAN: Deep Light Enhancement without Paired Supervision uses a deep learning algorithm that does not require paired supervision, which means that it can improve images without having access to corresponding high-quality images for comparison. Instead, it uses an adversarial training approach that pits two neural networks against each other: a generator network that enhances the images and a discriminator network that tries to distinguish between the enhanced images and the original ones.
The authors of this paper assemble a mixture of 914 low light and 1016 normal light images from several existing datasets and HDR sources, without the need to keep any pair. Manual inspection and selection are performed to remove images of medium brightness. All these photos are converted to PNG format and resized to 600×400 pixels.

EnlightenGAN-Dataset
Artifact overview
Type
dataset
Created At
January 26th, 2023
Description
Versions
Version
Aliases
Logged By
Tags
Created
TTL Remaining
# of Consuming Runs
Size
m.GitHub
m.Google-Drive
1
latest
v1
Thu Jan 26 2023
Inactive
2
9.7GB
https://github.com/csjcai/SICE
0
v0
Thu Jan 26 2023
Inactive
7
815.4MB
-
https://drive.google.com/drive/folders/1fwqz8-RnTfxgIIkebFG2Ej3jQFsYECh0?usp=sharing
Loading...

Citation

@misc{
1906.06972,
Author = {Yifan Jiang and Xinyu Gong and Ding Liu and Yu Cheng and Chen Fang and Xiaohui Shen and Jianchao Yang and Pan Zhou and Zhangyang Wang},
Title = {EnlightenGAN: Deep Light Enhancement without Paired Supervision},
Year = {2019},
Eprint = {arXiv:1906.06972},
}



DARK FACE: Face Detection in Low Light Condition

The DARK FACE dataset provides 6000 real-world low light images captured during the nighttime, at teaching buildings, streets, bridges, overpasses, parks, etc., all labeled with bounding boxes for human faces, as the main training/validation sets. The dataset also provides 9000 unlabeled low-light images collected from the same setting. Additionally, it provides a unique set of 789 paired low-light/normal-light images captured in controllable real lighting conditions (but unnecessarily containing faces), which can be used as parts of the training data at the participants' discretization. There will be a hold-out testing set of 4000 low-light images, with human face bounding boxes annotated.

DarkFace-Dataset
Artifact overview
Type
dataset
Created At
March 30th, 2023
Description
Versions
Version
Aliases
Logged By
Tags
Created
TTL Remaining
# of Consuming Runs
Size
m.source
0
latest
v0
Thu Mar 30 2023
Inactive
4
4.4GB
Loading...

Citation

@ARTICLE{
poor_visibility_benchmark,
author={
Yang, Wenhan and Yuan, Ye and Ren, Wenqi and Liu, Jiaying and Scheirer, Walter J. and Wang, Zhangyang and Zhang, and et al.
},
journal={IEEE Transactions on Image Processing},
title={
Advancing Image Understanding in Poor Visibility Environments: A Collective Benchmark Study
},
year={2020},
volume={29},
number={},
pages={5737-5752},
doi={10.1109/TIP.2020.2981922}
}

@inproceedings{
Chen2018Retinex,
title={Deep Retinex Decomposition for Low-Light Enhancement},
author={Chen Wei, Wenjing Wang, Wenhan Yang, Jiaying Liu},
booktitle={British Machine Vision Conference},
year={2018},
}



LLIV-Phone

The LLIV-Phone dataset contains 120 videos (45148 images) taken by 18 different phone cameras including iPhone 6s, iPhone 7, iPhone7 Plus, iPhone8 Plus, iPhone 11, iPhone 11 Pro, iPhone XS, iPhone XR, iPhone SE, Xiaomi Mi 9, Xiaomi Mi Mix 3, Pixel 3, Pixel 4, Oppo R17, Vivo Nex, LG M322, OnePlus 5T, Huawei Mate 20 Pro under diverse illumination conditions (e.g., weak illumination, underexposure, dark, extremely dark, back-lit, non-uniform light, color light sources, etc.) in the indoor and outdoor scenes.

LLIV-Phone
Artifact overview
Type
dataset
Created At
March 31st, 2023
Description
Versions
Version
Aliases
Logged By
Tags
Created
TTL Remaining
# of Consuming Runs
Size
m.source
0
latest
v0
Fri Mar 31 2023
Inactive
2
4.2GB
Paper: https://arxiv.org/pdf/2104.10729.pdf GitHub: https://github.com/Li-Chongyi/Lighting-the-Darkness-in-the-Deep-Learning-Era-Open Google-Drive: https://drive.google.com/file/d/1QS4FgT5aTQNYy-eHZ_A89rLoZgx_iysR/view
Loading...

Citation

@article{
2104.10729,
Author = {Chongyi Li and Chunle Guo and Linghao Han and Jun Jiang and Ming-Ming Cheng and Jinwei Gu and Chen Change Loy},
Title = {Low-Light Image and Video Enhancement Using Deep Learning: A Survey},
Year = {2021},
Eprint = {arXiv:2104.10729},
Howpublished = {TPAMI 2021},
}



Unpaired Low-Light Dataset

The Unpaired Low-Light Dataset is a dataset that is composed of images captured in diverse lighting conditions from various cameras compiled from all the aforementioned datasets. This dataset is intended to be used for training low-light enhancement models that don't rely on paired supervision.
The following Weave panel shows the Weights & Biases artifact corresponding to the dataset. You can check the lineage of this artifact to check how this dataset was compiled.

unpaired-low-light-dataset
Artifact overview
Type
dataset
Created At
March 31st, 2023
Description
Versions
Version
Aliases
Logged By
Tags
Created
TTL Remaining
# of Consuming Runs
Size
m.datasets
4
latest
v4
Fri Mar 31 2023
Inactive
1
2.1GB
Panel currently unsupported for custom step metrics. Please use an expression that returns the pre-defined step metric (_step) for now
3
v3
Fri Mar 31 2023
Inactive
3
4.4GB
Panel currently unsupported for custom step metrics. Please use an expression that returns the pre-defined step metric (_step) for now
2
v2
Fri Mar 31 2023
Inactive
1
623.8MB
Panel currently unsupported for custom step metrics. Please use an expression that returns the pre-defined step metric (_step) for now
1
v1
Fri Mar 31 2023
Inactive
1
965.0MB
Panel currently unsupported for custom step metrics. Please use an expression that returns the pre-defined step metric (_step) for now
0
v0
Fri Mar 31 2023
Inactive
1
348.1MB
Panel currently unsupported for custom step metrics. Please use an expression that returns the pre-defined step metric (_step) for now
Loading...


artifact
artifact
artifact
artifact
artifact
artifact