EdgeAI Studio Radar Model Composer Backend

This repository contains the source code for Radar's Model Composer Extension, also called Edge AI Studio, found on:
https://dev.ti.com/edgeaistudio/
and
https://www.ti.com/tool/download/EDGE-AI-STUDIO-MCU

Usage

The intended usage of this repository is to run the EdgeAI Studio program without the front-end GUI. This gives visibility to the backend source code which allows for modification and scripting of the existing functions.

Requirements

Python 3.10 package requirements:
pip install scikit-learn
pip install torch torchvision
pip install torchmetrics
pip install --upgrade git+https://github.com/ThanatosShinji/onnx-tool.git
pip install pandas
pip install --extra-index-url https://software-dl.ti.com/mctools/esd/tvm/mcu ti_mcu_nnc
pip install --upgrade --force-reinstall onnx==1.13.0

Running

To run the backend, demo.py must be called properly. The method to run demo.py with no changes needed can be found in example_cmd_runs.txt. When all the prerequiste Python packages have been installed, using the notation shown in this file will allow you to run the full extension from dataset importing all the way to generating a model optimized as C code.

If Edge AI Studio was never installed locally, the paths seen for input_data_path, train_output_path, and compile_output_path will not exist. You are able to change these freely when running the extension directly from demo.py

Need More Help?

• Find the Edge AI Studio user guide for Radar in the Radar Toolbox
• Learn about Radar fundamentals and types of data available using modules found in our Radar Academy
• For more details on the TI's TVM Compiler, checkout the TI NNC User's Guide
• Search for your issue or post a new question on the mmWave E2E forum