TinyML Talks - Speech-to-Intent Model Deployment to Low-Power Low-Footprint Devices

Explore speech-to-intent model deployment techniques for low-power, low-footprint devices in this tinyML Talk. Discover efficient methods for parsing user utterances directly into actionable output, bypassing traditional speech-to-text transcription. Learn to train domain-specific models and deploy them on Cortex M4F-based development boards with built-in microphones, such as the Wio Terminal from Seeed Studio. Gain insights into data set creation, model training, and accuracy improvements for resource-constrained microcontrollers. Witness a live demonstration and explore potential applications across various targets. Delve into challenges, solutions, and future improvements in this cutting-edge field of tinyML.

Introduction
Welcome
Types of speech processing
Keyword spotting
Large vocabulary
Comparison
A better way
About the device
Data set
Results
Accuracy
Questions
Speechtointent model
Challenges
Data sets
Wire Terminal
Demo
Installing dependencies
Test data set
Baseline model
Training
Demonstration
Possible targets
Improvements
Fork code repository
Final words
Virtual machine
Sponsors