A structured ensemble of Binarized Neural Networks for image classification on Raspberry Pi Pico.
This project is an attempt to deploy on Raspberry Pi Pico a promising structured ensemble of Binarized Neural Networks (BNNs) called BeMi Stardust. The authors trained BNNs with a Mixed Integer Programming (MIP) model, however in this implementation they were trained with classical machine learning techniques (e.g. Adam). This work is an extention of my bachelor's thesis in Bioengineering at Università degli Studi di Pavia. You can find out more here.
- LowBeMi notebook: contains everything you need in order to train all the BNNs of the ensemble using the Larq library.
- More coming out soon ...
LowBeMi is available under the MIT Licence. Use it wisely.