This is Tripod, an AI Image Sharpener. It tries its best to remove or reduce motion blur from pictures.
Feel free to test it, use it, improve it.
Right now, exetremely minimal usecases are covered:
- Tripod is only meant to receive as input square images with sizes that are powers of 2 (so 256x256, 512x512, 1024x1024). Crop your images accordingly before using it, or be prepared to read error logs.
- Tripod was trained on images that are artificially blurred. The guy is trying his best, don't expect him to perfectly sharpen a real-world blurry image.
git clone https://github.com/giorgioskij/tripod
cd tripod
conda env create -f environment.yml
conda activate tripod
The best way to use Tripod is to interact with the code directly. However, if you just want to test the model, you can run
python scripts/demo.py <path to your image>
The sharpened image should be saved in the same directory as the input one.
Here follow a few examples of comparing blurred input, sharpened output, and sharp original.
