lowlight enhance mirnet

Introduction

The Low-light Image Enhancement using MIRNet repository contains a model and a notebook designed to improve the quality of images captured under low-light conditions. The model, a fully-convolutional architecture, is capable of learning enriched features that combine contextual information from multiple scales while maintaining high-resolution spatial details. This work has been reproduced by Vu Minh Chien with slight modifications to the hyperparameters, leveraging the original work by Soumik Rakshit.

Architecture

The MIRNet model is a fully-convolutional architecture designed for image restoration tasks, especially in low-light conditions. It effectively combines contextual information from multiple scales and preserves high-resolution spatial details, making it suitable for applications in photography, security, medical imaging, and remote sensing.

Training

The training utilizes the LoL Dataset, which consists of 485 images for training and 15 for testing. Each image pair includes a low-light input and a corresponding well-exposed reference image. The training was conducted with the following hyperparameters:

• learning_rate: 1e-04
• train_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9, 0.999) and epsilon=1e-08
• lr_scheduler_type: ReduceLROnPlateau
• num_epochs: 50

The results can be viewed through TensorBoard, where training metrics are detailed.

Guide: Running Locally

1. Clone the Repository: Use Git to clone the repository to your local machine.

   git clone https://github.com/keras-io/lowlight-enhance-mirnet.git
   

2. Install Dependencies: Navigate to the project directory and install the required Python packages.

   cd lowlight-enhance-mirnet
   pip install -r requirements.txt
   

3. Download Dataset: Obtain the LoL Dataset from the provided link and place it in the appropriate directory.
4. Run the Model: Execute the training script or notebook for the model.

   python train.py
   

For optimal performance, consider using cloud GPUs such as those provided by AWS, Google Cloud, or Azure.

License

The project does not specify a license in the provided documentation. Always check the repository for any licensing information before use.