TinyML-Examples
In this repository you will find a list of submodules that correspond to the following Tiny(M)achine(L)earning examples.
Examples' status
| Example | Status | Contact |
|---|---|---|
| Sine (Hello World) |  & |
Alex |
| MNIST | & |
Alex/Philip |
| MicroSpeech | & |
Alex/Philipp |
| CIFAR10 |  |
- |
| CNN |  |
Yu-Kai |
| Graph | |
- |
| Face Recognition | |
Alex/Teo |
| KWS | |
- |
Documentation/Notes
Sine/Hello World
The colab notebook to generate the model is directly taken from the book TinyML and the working implementation is for a STM3240G-EVAL development board.
Working implementation can be found here.
Offline interpreter example can be found on the compiler branch of the implementation repository.
MNIST
The MNIST example is currently bottlenecked by a large RAM requirement, probably im2col.
Practical implementation is being constructed here.
MicroSpeech
Wake word example from TinyML book. FP of Philipp van Kempen.
Practical implementation for STM32F769 board.
CIFAR10
CNN
Graph
Face Recognition
Bachelor thesis topic of Teo Fratiloiu.
Building a data set was part of Teo's IP and led to the development of the webscrapper which aims to automate the scrapping, validation and preprocessing of image datasets.
Practical implementation is being constructed here.