LauzHack Deep Learning Bootcamp
-
Updated
Jul 19, 2025 - Jupyter Notebook
LauzHack Deep Learning Bootcamp
On-Device Learning for Human Activity Recognition on Low-Power Microcontrollers
Code for "A Federated Approach for Adaptive Urban Sound Classification on TinyML Edge Devices" (Sensors 2026, 26, 2854). Federated on-device learning on ESP32, a 231-parameter head trains locally and syncs over MQTT at 1.2 kB per round.
An anomaly detector that trains on the machine it watches. Unsupervised on-device learning in fixed point on an ESP32 - no labels, no cloud, no pre-trained fault classes.
AegisFL is a cloud-native, privacy-preserving federated learning platform. It uses TensorFlow Federated, Differential Privacy, and Secure Aggregation to train models across decentralized clients, ensuring HIPAA/GDPR compliance with cost-optimized Kubernetes deployment and real-time monitoring.
To associate your repository with the on-device-learning topic, visit your repo's landing page and select "manage topics."