Part of my Engineering Science (ECE option) degree at the University of Toronto. Supervised by Prof. Xilin Liu of UofT X-Lab.
Project description: https://tristan.cbcr.me/?type=a#edge_AI
Install Python dependencies with: pip3 install -r python_prototype/requirements.txt.
Install EdgeTPU runtime and PyCoral by following this guide.
To generate the Makefile, CMake is required. Install it with brew install cmake.
To compile the simulation, gcc, hdf5, boost and armadillo are required. Install them with brew install gcc hdf5 boost armadillo.
Install verilator for Verilog compilation (requires MacOS Homebrew): brew install verilator.
Install gtkwave to view HDL simulation waveforms (requires MacOS Homebrew): brew install --cask gtkwave.
Install Python dependencies for IP verification with pip3 install -r asic/ip_verif/requirements.txt
To compile LaTeX documents, install latexml, texlive and biber: brew install latexml texlive biber.
You'll also need the Python package SciencePlots to export plots used in the thesis. Install it with: python -m pip install git+https://github.com/garrettj403/SciencePlots.git
I recommend using the LaTeX Workshop extension in VSCode.
In this repository, you will find most files for this thesis.
In /python_prototype, you'll find all files for the vision transformer model, written using TensorFlow. There all also files for signal extract and data processing from raw polysomnography datasets. Finally, it contains shell scripts to train on the Compute Canada cluster and some utilities files.
In /latex, you'll find all deliverables for this course (ESC499).
In /asic, you'll find a C++ functional model of the ASIC accelerator and most files for its synthesis (SystemVerilog, testbenches, etc.). Some files are excluded as they may contain information under NDA.