Recognizing MNIST-based Handwritten Digits on M5Stack Core2

 

In his previous tutorial, Sumit had great fun adding handmade drawing gestures for the M5Stack Touch Screen device and managed to prove that a low-power MCU can classify complex gestures in real-time at the edge using the TinyML technology. 

Today, we are sharing another experiment by Sumit Kumar, where he achieved the same for handwritten digit recognition similar to MNIST (since MNIST is so popular due to its size, which allows deep learning researchers to quickly check and prototype their algorithms).

In this project, Sumit will demonstrate how to build a TinyML model for recognizing handwritten digits on touch interfaces for simple MCUs, and provide detailed guidelines on how to make your own dataset, train the model on a no-code platform, and run the model on the device (video included).

Read the full version on hackster.io