What is Tiny Machine Learning?
Deep learning and embedded systems combine in tiny devices
Tiny Machine Learning (TinyML) is the latest embedded software technology is about making computing at the edge cheaper, less expensive and more predictable.
TinyML is to some extent about how to best implement machine learning (ML) in ultra-low power systems.
There are two aspects in this shift that may make this easy manage:
- Arduino, known for open-source hardware is making TinyML available for millions of developers.
- Edge Impulse turning Arduino into an embedded ML platform.
Tiny sensors along with low-powered microcontrollers.
TinyML is said to be the missing link between edge hardware and device intelligence.
An article in TechCrunch from the 29th of May calls it:
“Tiny devices with not-so-tiny brains”
There is a book from the publisher O’Reilly called TinyML released in 2019.
It argues that deep learning networks are getting smaller.
“The Google Assistant team can detect words with a model just 14 kilobytes in size — small enough to run on a microcontroller.”
Deep learning and embedded systems combine in tiny devices.
It is possible to train models small enough to fit into most environments.
It is possible to build a speech recogniser, with a camera that detects people, and a response. Arduino, ultra-low-power microcontrollers together with TensorFlow Lite, and Google’s toolkit can be combined.
Rather amazing how small, connected devices increasingly can process compute for these new tasks.
This is #500daysofAI and you are reading article 361. I am writing one new article about or related to artificial intelligence every day for 500 days. My focus for day 300–400 is about AI, hardware and the climate crisis.
