This tutorial is designed to teach you how to create a gesture detection AI model that uses a Radar sensor. This can be used for a wide variety of applications such as control inputs for a wireless headset.
This tutorial will cover the whole process from the data collection phase through to generating the edge source code.
# Software Versions Used
This section aims to include the version of different software used for reader awareness. This should be kept in mind if images, functionality does not reflect that documented in this tutorial.
# Imagimob Studio
Version 2.6.158 BETA. Imagimob Studio is used for data labelling, model building and model evaluation.
# ImagUtils (Optional)
Master branch - cd1d189. ImagUtils is used for auxiliary functionality such as automatic labelling of data.
# Capture Server
Master branch - 1cb0577. Capture Server is used for connecting to the sensor, receiving data and synchronising data with a video that is recorded.
# Acconeer Radar Firmware
Version 2.7.2. Acconeer_xm122_module_software is used for flashing the XB112/XB122 board.
# Project Download
The project can be downloaded inside of Imagimob Studio by clicking on Create New Project and selecting the Acconeer Radar Gesture Project. After clicking on OK, the download process starts. Once it is done, you will get a folder that includes
- Data - Folder where data is located along with the relevant configuration file so that the data can be reproduced.
- IM Units - Folder where custom layers and pre-processors can be added and a tool to perform real time testing.
- Project Iteration # - Folder where the project file is located. All results related to a specific project file should be placed in the subfolder Results.
- Tools & Scripts - Folder where all extra tools and scripts belonging to the project should be placed. Here, we include the Acconeer Live Test Tool, which is used to perform live predictions on incoming data from a serial connection to the Acconeer radar.
This tutorial will be run with the Acconeer radar development board and this can be paired with the Imagimob Capture Server & App that allow you to read in serial data from the radar and transmit it over WiFi to the capture app for recording.
You'll need the following hardware and software:
- XM122 IoT MODULE (opens new window)
- Micro USB Cable
- Imagimob Capture Server (opens new window)
- Imagimob Capture Application (opens new window)
# Tutorial Index
This tutorial will be set up into the following parts. The primary idea is that each part represents a milestone of some sort regarding setting up AI/ML (artificial intelligence or machine learning) projects.
- Setting Up Capture System
- Data Collection
- Data Labelling & Management
- Label Analysis & Distribution
- Model Building & Training
- Model Evaluation
- Edge Optimization
- Edge Code & Deployment
- Closing Comments
# Acconeer Documentation
This tutorial focuses on the model building part of the project. To find more information on how the radar works, the best way to use it and the possibility of deploying your model on an Acconeer dev kit please visit Acconeer's developer pages. Acconeer has extensive documentation on their own website found at the following links:
- developer.acconeer.com (opens new window)
- acconeer-python-exploration.readthedocs.io (opens new window)