Basic Python based VQ Autoencoder - Tested on CocoData
Adapted from functionally programmed code created during model based RL research in 2024. This specific VQAE is intended for use in edge-case computing with a focus on good reconstruction quality with the expectation of limited processing and input resolution.
Main tested metrics are reconstruction loss, PSNR and SSIM
Pre-trained model files trained on the 2017 CocoData dataset