tinyML Talks: Unpacking the music genre recognition project from the TinyML Cookbook, second edition

"Unpacking the music genre recognition project from the TinyML Cookbook, second edition!"
Gian Marco Iodice
Team and Tech Lead in the Machine Learning Group
Arm
In this exclusive practice session, Gian Marco Iodice will demonstrate how to build a music genre recognition application on the Raspberry Pi Pico using TensorFlow Lite for Microcontrollers and the CMSIS-DSP library. This project is part of the TinyML Cookbook's second edition, and it is proposed to demonstrate how the target device influences our design choice, from the feature extraction to the model design, when deploying ML models on microcontrollers.
The talk will start by tailoring the Mel-frequency cepstral coefficients (MFCCs) feature extraction for the Raspberry Pi Pico. Here, you will learn how fixed-point arithmetic can help minimize the latency performance and show how the CMSIS-DSP library provides tremendous support in employing this numerical format.
Afterward, Gian Marco will present the design choices for the ML model capable of recognizing music genres with a long-short-term memory (LSTM) recurrent neural network (RNN).
Finally, he will show how to deploy the final application on the Raspberry Pi Pico with the help of TensorFlow Lite for Microcontrollers.
A book giveaway will follow at the end of this presentation for the chance to get a free copy of the second edition of the TinyML Cookbook!