Hello everyone,
I would like to share with you an implementation of a ResNet20 to classify encrypted images using the OpenFHE implementation of CKKS.
The algorithm requires at most 16GB of RAM, and is freely accessible as a GitHub repository. I also share an article on ePrint that explains the used approach, aimed to decrease the high memory requirements of CKKS by minimizing the number of rotation keys.
I would not be surprised in case of mistakes, I am still learning so feel free to try it out and report bugs or issues!
The circuit has been built and tested on v.1.0.4
Thank you