Low-Light Image Enhancement using Convolutional Neural Networks

This project implements a Convolutional Neural Network (CNN) for enhancing low-light images. The model is designed to improve the visual quality of images captured in low-light conditions. This README provides an overview of the project, including setup instructions, usage, and additional information.

Table of Contents

• Introduction
• Project Structure
• Setup Instructions
• Data Preparation
• Training the Model
• Evaluation
• Usage
• Contributing
• License

Introduction

Low-light image enhancement is a crucial task in computer vision, aiming to improve the quality of images taken in poor lighting conditions. This project leverages a CNN model to enhance low-light images by learning the mapping from low-light images to their enhanced versions.

Project Structure

• data/: Directory containing training and validation datasets.
• src/: Source code for the model, training script, and evaluation script.
• notebooks/: Jupyter notebooks for data exploration and experimentation.
• README.md: This README file.
• requirements.txt: Required Python packages.

Setup Instructions

1. Clone the repository:

   git clone https://github.com/your-username/low-light-image-enhancement.git
   cd low-light-image-enhancement

2. Set up a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install the required packages:

   pip install -r requirements.txt

4. Mount Google Drive (if using Google Colab):

   from google.colab import drive
   drive.mount('/content/drive')

Data Preparation

1. Load your data:

   • Ensure your datasets are stored in the appropriate directories (data/train/low_light, data/train/enhanced, data/val/low_light, data/val/enhanced).

2. Convert lists to numpy arrays (if needed):

   import numpy as np
   low_light_images_list = np.load('/path/to/low_light_images.npy', allow_pickle=True)
   enhanced_images_list = np.load('/path/to/enhanced_images.npy', allow_pickle=True)
   low_light_images = np.array(low_light_images_list)
   enhanced_images = np.array(enhanced_images_list)

Training the Model

1. Run the training script:

   python src/train.py

   This script will train the CNN model using the specified dataset and hyperparameters. Gradient accumulation is implemented to handle larger batch sizes.

Evaluation

1. Evaluate the model:

   python src/evaluate.py

   This script will evaluate the model's performance using the Peak Signal-to-Noise Ratio (PSNR) metric on the validation dataset.

Usage

1. Enhance a new image:

   Use the trained model to enhance new low-light images. Ensure the model is loaded correctly and process the images as follows:

   from src.model import build_enhanced_model
   import numpy as np
   import cv2
   
   # Load the model
   model = build_enhanced_model(input_shape)
   model.load_weights('path/to/model/weights')
   
   # Load and preprocess the low-light image
   low_light_image = cv2.imread('path/to/low_light_image.png')
   low_light_image = low_light_image / 255.0
   low_light_image = np.expand_dims(low_light_image, axis=0)
   
   # Enhance the image
   enhanced_image = model.predict(low_light_image)
   enhanced_image = np.squeeze(enhanced_image, axis=0)
   
   # Save the enhanced image
   cv2.imwrite('path/to/enhanced_image.png', enhanced_image * 255)

Contributing

Contributions are welcome! Please read the contributing guidelines for more information on how to contribute to this project.

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

This README file provides a comprehensive guide for setting up, using, and contributing to the low-light image enhancement project. Follow the instructions carefully to ensure proper setup and execution of the project. For any issues or questions, feel free to open an issue on GitHub.