This repository hosts the code for TinyissimoYOLO and DSORT-MCU as presented in Ultra-Efficient On-Device Object Detection on AI-Integrated Smart Glasses with TinyissimoYOLO and DSORT-MCU: Detecting Small Objects in Real Time on MCUs.

👓 Ultra-Efficient On-Device Object Detection on AI-Integrated Smart Glasses with TinyissimoYOLO

Ultra-Efficient On-Device Object Detection on AI-Integrated Smart Glasses with TinyissimoYOLO
Julian Moosmann* ¹, Pietro Bonazzi*¹, Yawei Li¹, Sizhen Bian ¹, Philipp Mayer¹ , Luca Benini ¹ , Michele Magno¹

¹ ETH Zurich, Switzerland

Guide

For more details on how to train TinyissimoYOLO models, follow this guide.

DSORT-MCU: Detecting Small Objects in Real Time on MCUs

DSORT-MCU is a framework for training and running models with improved detection performance on small objects that does not increase memory footprint and thus enables detection on resource constrained platforms.

DSORT-MCU: Detecting Small Objects in Real Time on MCUs

Liam Boyle ¹, Julian Moosmann¹, Nicolas Baumann ¹, Seonyeong Heo² , Michele Magno¹

¹ ETH Zurich, Switzerland
² Kyung Hee University, Republic of Korea

Guide

For a guide on how to train models with the adaptive tiling presented in DSORT-MCU follow these instructions.