Honey, I shrunk the TinyML | Lars Gregori

Python

Discover how TinyML enables machine learning on microcontrollers, reducing complexity and size to run locally without cloud processing, using the Raspberry Pi Pico as a demonstration.

Key takeaways

The talk revolves around TinyML, a technology that enables machine learning to run on microcontrollers, which is small, low-power, and low-cost.
The Raspberry Pi Pico is a good example of a microcontroller that can run TinyML, having an ARM Cortex M0+ processor and 264KB of RAM.
The first example shown is an XOR model, which is a simple machine learning model that can be implemented on the Raspberry Pi Pico.
The Pico is programmed using MicroPython, a variant of Python that is optimized for microcontrollers.
The TensorFlow Lite interpreter is used to convert the machine learning model into a binary that can be run on the microcontroller.
The key advantage of TinyML is that it allows machine learning to be run locally on the microcontroller, without the need for an internet connection or cloud processing.
The talk highlights the importance of reducing the size and complexity of machine learning models to enable them to run on microcontrollers.
The Pico has a limited amount of memory and processing power, which requires the model to be optimized and quantized to run efficiently.
The optimization process involves reducing the number of layers and nodes in the model, and using quantization to reduce the precision of the model’s weights and biases.
The talk concludes that TinyML is a promising technology that has the potential to enable machine learning to be used in a wide range of applications, from home automation to wearables to Internet of Things (IoT) devices.

Honey, I shrunk the TinyML | Lars Gregori

More talks