Review Article

Tiny Machine Learning for Resource-Constrained Microcontrollers

Table 2

TinyML development tools’ features.

Development toolAlgorithms and supportTarget processor

Edge Impulse StudioProprietary NNs32-bit (Cortex-M0+–M7)
Qeexo AutoMLGBM, RF, XGBoost, NB, DT, IF, LR, LOF, SVM, CNN, CRNN, RNN, ANN32-bit (Cortex-M0–M4)
STM32Cube.AI (+X-CUBE-AI)Scikit-learn (IF, SVM, kMC, etc.), XGBoost, Keras, TFLite, Caffe, Lasagne, ConvnetJS32-bit (Cortex-M0⟶)
NanoEdge AI StudioNN, kNN, SVM, proprietary ML32-bit (Cortex-M0⟶)
ImagimobProprietary NNs32-bit (Cortex-M0⟶)
emlearnRF, DT, NB, MLP, Keras, scikit-learn8-bit (ATmega328P⟶)