MAXIM pre-trained on LOL for image enhancement

MAXIM model pre-trained for image enhancement. It was introduced in the paper MAXIM: Multi-Axis MLP for Image Processing by Zhengzhong Tu, Hossein Talebi, Han Zhang, Feng Yang, Peyman Milanfar, Alan Bovik, Yinxiao Li and first released in this repository.

Disclaimer: The team releasing MAXIM did not write a model card for this model so this model card has been written by the Hugging Face team.

Model description

MAXIM introduces a shared MLP-based backbone for different image processing tasks such as image deblurring, deraining, denoising, dehazing, low-light image enhancement, and retouching. The following figure depicts the main components of MAXIM:

Training procedure and results

The authors didn't release the training code. For more details on how the model was trained, refer to the original paper.

As per the table, the model achieves a PSNR of 23.43 and an SSIM of 0.863.

Intended uses & limitations

You can use the raw model for image enhancement tasks.

The model is officially released in JAX. It was ported to TensorFlow in this repository.

How to use

Here is how to use this model:

from huggingface_hub import from_pretrained_keras
from PIL import Image

import tensorflow as tf
import numpy as np
import requests

url = "https://github.com/sayakpaul/maxim-tf/raw/main/images/Enhancement/input/748.png"
image = Image.open(requests.get(url, stream=True).raw)
image = np.array(image)
image = tf.convert_to_tensor(image)
image = tf.image.resize(image, (256, 256))

model = from_pretrained_keras("google/maxim-s2-enhancement-lol")
predictions = model.predict(tf.expand_dims(image, 0))

For a more elaborate prediction pipeline, refer to this Colab Notebook.

Citation

@article{tu2022maxim,
  title={MAXIM: Multi-Axis MLP for Image Processing},
  author={Tu, Zhengzhong and Talebi, Hossein and Zhang, Han and Yang, Feng and Milanfar, Peyman and Bovik, Alan and Li, Yinxiao},
  journal={CVPR},
  year={2022},
}