Embedded ML System

Open Challenge

We all know the benefits of using Machine Learning and Artificial Intelligence to automate and resolve tasks to reduce cost. However deploying ML/AI would require massive hardware which may not be deployed depending on the location (such as in carparks,  farms, etc). So how do we bring these capabilities if hardware becomes an issue?


How can one utilize the power of ML/AI in a situation where:
  • Power Supply is limited
  • Internet Connection is scarce
  • Needs to operate in areas where weather conditions are harsh
By integrating the power of embedded systems!


We deployed a Machine Learning model on an embedded system which is not only small but also robust. It requires no internet connection and can run on batteries for months at a time. With proper casing, it can be weather-proofed as well. Here the embedded system is designed to work with a camera to act as a person detection module:

Some of the functions that can be built-in are such as Text Recognition, Human Recognition, Movement Detection and Object Semantics.

Proposed Architecture