ML-TN-008 — AI at the edge: prototyping an IoT real-time endoscope

Info Box Applies to Machine Learning Applies to Machine Learning Technical Notes

History

Version	Date	Notes
0.1.0	March 2025	First public draft

Abstract

This Technical Note describes Luca Manzini's activities during his internship at DAVE Embedded Systems. By using an embedded platform, he prototyped an IoT smart endoscope for automatic detection of gastro-intestinal tract polyps. This device makes use of a deep-learning model to perform real-time

Major improvements in the field of medicine, particularly in the diagnosis and treatment of numerous diseases, have been made possible by the progress of technology, Artificial Intelligence (AI) in particular. Among these is the real-time identification of intestinal polyps, which can be considerably enhanced by computer vision and embedded devices. By leveraging an embedded platform design for industrial applications, the main goal of the internship was to implement a system capable of detecting intestinal polyps in real time thanks to deep-learning-based, FPGA-hardware-accelerated inference algorithms. The suggested solution takes advantage of the well-known and effective YOLO (You Only Look Once) architecture for object detection in images. PyTorch, an open-source deep learning framework with a large feature set and a multitude of tools, is used to construct this architecture.