qualia-core 2.0.0

Qualia toolchain Core

Qualia Core (formerly MicroAI)

End-to-end training, quantization and deployment framework for deep neural networks on microcontrollers.

Repository should be cloned with --recursive to get TFLite Micro and its dependencies.

Dependencies

Python:

numpy
scikit-learn
tomlkit
colorful
gitpython

Dataset

GTSRB

Python:

imageio
scikit-image

Training

TensorFlow

Python:

tensorflow
tensorflow_addons

PyTorch

Python:

pytorch
pytorch_lightning

Deployment

Embedded targets

SparkFun Edge

Python:

pycryptodome

Nucleo-L452RE-P

System:

stm32cubeide
stm32cubeprog

Embedded frameworks

STM32Cube.AI

STM32CubeIDE extension pack:

X-CUBE-AI == 5.2.0

TensorFlow Lite Micro

System:

arm-none-eabi-binutils
arm-none-eabi-gcc
arm-none-eabi-newlib
libopenexr-dev
wget

Qualia-CodeGen

Python:

jinja2

System:

arm-none-eabi-binutils
arm-none-eabi-gcc
arm-none-eabi-newlib

Evaluation

Python:

pyserial

Usage

If Qualia installed with pip, you can run the qualia command directly. Otherwise run PYTHONPATH=. ./bin/qualia <config.toml> <action> from the qualia directory.

Dataset pre-processing

qualia <config.toml> preprocess_data

Training

qualia <config.toml> train

Prepare deployment (generate firmware)

qualia <config.toml> prepare_deploy

Deploy and evaluate

qualia <config.toml> deploy_and_evaluate

Run test suite

CUBLAS_WORKSPACE_CONFIG=:4096:8 PYTHONHASHSEED=2 python -m unittest discover qualia/tests

Included support for datasets, learning framework, neural networks, embedded frameworks and targets

Datasets

• UCI HAR
• GTSRB
• Written and Spoken MNIST

Learning frameworks

• TensorFlow.Keras
• PyTorch

Neural networks

• MLP
• CNN (1D&2D)
• Resnetv1 (1D&2D)

Embedded frameworks

• STM32Cube.AI
• TensorFlow Lite for Microcontrollers
• Qualia-CodeGen

Targets

• Nucleo-L452RE-P
• SparkFun Edge

Reference & Citation

Quantization and Deployment of Deep Neural Networks on Microcontrollers, Pierre-Emmanuel Novac, Ghouthi Boukli Hacene, Alain Pegatoquet, Benoît Miramond and Vincent Gripon, Sensors, 2021.

@article{qualia,
	author = {Novac, Pierre-Emmanuel and Boukli Hacene, Ghouthi and Pegatoquet, Alain and Miramond, Benoît and Gripon, Vincent},
	title = {Quantization and Deployment of Deep Neural Networks on Microcontrollers},
	journal = {Sensors},
	volume = {21},
	year = {2021},
	number = {9},
	article-number = {2984},
	url = {https://www.mdpi.com/1424-8220/21/9/2984},
	issn = {1424-8220},
	doi = {10.3390/s21092984}
}