Low-Power Anomaly Detection and Classification System based on a Partially Binarized Autoencoder for In-Sensor Computing | IEEE Conference Publication | IEEE Xplore