This is an official pytorch implementation of our paper: "STMemAE: An Instance-Level based Spatio-Temporal Memory Autoencoder for Unsupervised Vision-based Seizure Detection", which is currently under review. In this repository, we provide PyTorch code for only the proposed STMemAE model. STMemAE provides an unsupervised vision based seizure detection framework that achieves high performance on CHZU video epileptic dataset. The rest of the codes will be published after our paper is accepted！