Why Machine Learning on Microcontrollers?
summary
uTensor takes TensorFlow graphs and generates C++ that runs on tiny chips and the demo shows MNIST handwriting and keyword spotting running on a Cortex M4 so it actually works not just promising ideas. The claim includes energy math: one image every five seconds on a single coin cell for >1 year. It also shows sensor fusion and edge anomaly detection, describes a workflow and graph optimizations, and reports real deployments and bandwidth savings.
AI-written summary · about
chapters (7)
related talks
25mMachine Learning on Microcontrollers; Tiny, Useful and Fun
27mWhat’s TinyML good for
01h 53mTinyML: Machine Learning’s Future is Tiny and Bright
31mKeynote: The next big thing in machine learning is tiny
24mRunning AI/Neural networks on microcontrollers made simple with the STM32Cube.AI
01h 01mML and the IoT: Living on the Edge
55mBuilding with TensorFlow Lite for microcontrollers
27mMachine Learning on the edge using TensorFlow Lite