Model description
This repo contains the model and the notebook Low-light image enhancement using MIRNet.
Full credits go to Soumik Rakshit
Reproduced by Vu Minh Chien with a slight change on hyperparameters.
With the goal of recovering high-quality image content from its degraded version, image restoration enjoys numerous applications, such as photography, security, medical imaging, and remote sensing. The MIRNet model for low-light image enhancement is a fully-convolutional architecture that learns an enriched set of features that combines contextual information from multiple scales, while simultaneously preserving the high-resolution spatial details
Dataset
The LoL Dataset has been created for low-light image enhancement. It provides 485 images for training and 15 for testing. Each image pair in the dataset consists of a low-light input image and its corresponding well-exposed reference image.
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- 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
Training results
- The results are shown in TensorBoard (Training metrics).
View Model Demo
<details>
<summary> View Model Plot </summary>
</details>
