The goal of this projet was to devleop methods for image cartoonization.
python = 3.8
You can find the report at this address: https://drive.google.com/drive/folders/1QZEOePZYLy5r-v1NzK5gERmB5CrUQt3P?usp=sharing.
All data is available at https://drive.google.com/drive/folders/1Ty-eD8pdJNja5iQjBSf8-8xe43IWF0ij?usp=sharing.
Style transfer weights: https://drive.google.com/drive/folders/1Ty-eD8pdJNja5iQjBSf8-8xe43IWF0ij.
To install dependencies please run make install.
To run demonstrator please run make demonstrator.
We developed a tab in the demonstrator to evaluate the results of the cartoonizer. Thanks to this tab, you'll be able to rank the different predictions to estimate which method giuves the best results.
Cartoonized images must be placed in the data/results/<model_id> where <model_id> is the folder containing all cartonnized images. All predicted images must have the same name as the original image. For example, for the test image 667626_18933d713e.jpg in flickr dataset should have the same name in the data\results\model_1 folder.