nul-image-enhancer's Introduction

The Non-Ultimate Image Enhancer 🐌

An algorithm based on decision trees and deep learning models to enhance images.
The algorithm is simple, it does not always provide good results on already great images.
The default maximum runtime is 60 seconds / image. The algorithm can provide good results in 10 seconds on some images.
The algorithm tries multiple transformations on contrast / saturation / lighting etc., it evaluates the changes and keeps the best ones.

➡️ More details on the algorithm: How I Built an AI Image Enhancer

Results

Enhanced ✔️

Enhanced ❓

Requirements

A GPU with > 2GB VRAM (very slow on CPU)
CLIP
Pytorch

Usage

python3 enhance_image.py examples/1.png --maxruntime=10

Available sd-webui extension: https://github.com/Whiax/sd-webui-nulienhance

Models

⭐ If you like it! Thanks!

nul-image-enhancer's People

Contributors

Stargazers

Watchers

nul-image-enhancer's Issues

module 'torchvision.transforms.functional' has no attribute 'adjust_sharpness'

I use python 3.7 environment. and also use python3 install.py to install environment.

But I exec python3 enhance_image.py examples/1.png --maxruntime=10 it shows:

Starting image enhancer...
- reading args...
- loading data...
- loading models...
- base_score: 5.392
Traceback (most recent call last):
  File "enhance_image.py", line 170, in <module>
    transform_to_params = get_transform_to_params(mode)
  File "/home/blue/projects/NUl-Image-Enhancer/models.py", line 59, in get_transform_to_params
    transform_to_params[FT.adjust_sharpness] = {'sharpness_factor':((a, 1/a),)}
AttributeError: module 'torchvision.transforms.functional' has no attribute 'adjust_sharpness'

Recommend Projects