MIRNet

Tensorflow implementation of the MIRNet architecture as proposed by Learning Enriched Features for Real Image Restoration and Enhancement.

Lanuch Notebooks:

Wandb Logs: https://wandb.ai/19soumik-rakshit96/mirnet

Blog Post: https://keras.io/examples/vision/mirnet/

TFLite Variant of MIRNet: https://github.com/sayakpaul/MIRNet-TFLite.

TFLite Models on Tensorflow Hub: https://tfhub.dev/sayakpaul/lite-model/mirnet-fixed/dr/1.

Tensorflow JS Variant of MIRNet: https://github.com/Rishit-dagli/MIRNet-TFJS.

Pre-trained Weights

• Trained on 128x128 patches: https://drive.google.com/file/d/1sUlRD5MTRKKGxtqyYDpTv7T3jOW6aVAL/view?usp=sharing

• Trained on 256x256 patches: https://drive.google.com/file/d/1sUlRD5MTRKKGxtqyYDpTv7T3jOW6aVAL/view?usp=sharing

Citation

@misc{
    2003.06792,
    Author = {Syed Waqas Zamir and Aditya Arora and Salman Khan and Munawar Hayat and Fahad Shahbaz Khan and Ming-Hsuan Yang and Ling Shao},
    Title = {Learning Enriched Features for Real Image Restoration and Enhancement},
    Year = {2020},
    Eprint = {arXiv:2003.06792},
}