Zero-Reference Deep Curve Estimation (ZeroDCE) for Low-Light Image Enhancement

PyTorch implementation of Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement Chongyi Li et al.

Execute code

• Copy the https link from GitHub repository.
• Using terminal use git clone <https link> in your desired directory.
• Type cd ZeroDCE in terminal.
• Execute <b>test_one.py</b> script to run pretrained model on a random image.
• Execute <b>train.py</b> to train a new model and save it in the models folder.

Main Model Architecture

Complete model which will iteratively apply pixel-wise transformations to an image to enhance it.

<p align="center"> <img src="Assets/main_model_architecture.png"> </p>

CNN Architecture

<p align="center"> <img src="Assets/CNN_model_architecture.png"> </p>

Loss Functions

• Spatial Consistency Loss
• Exposure Control Loss
• Color Constancy Loss
• Illumination Smoothness Loss

Prerequisites

• Pytorch
• NumPy
• python 3

Some Results

<p align="center"> <img src="Assets/result1.png"> </p> <p align="center"> <img src="Assets/result2.png"> </p> <p align="center"> <img src="Assets/result3.png"> </p> <p align="center"> <img src="Assets/result4.png"> </p>

Contributors:

• Ankur Chemburkar
• Devang Jain
• Hashir K