Computer vision could be a vital a part of ML apps of various scales, from $20,000 Tesla Bots or self-driving cars to smart doorbells and vacuum cleaners. It’s also a difficult task because, in comparison with a cloud infrastructure, on “real” edge devices, the hardware specs are sometimes far more constrained.
YOLO (You Only Look Once) is a well-liked object detection library; its first version was made in 2015. YOLO is especially interesting for embedded devices because it could actually run almost anywhere; there usually are not only Python but additionally C++ (ONNX and OpenVINO) and Rust versions available. A 12 months ago, I tested YOLO v8 on a Raspberry Pi 4. Nowadays, many things have modified — a brand new Raspberry Pi 5 became available, and a more recent YOLO v10 was released. So I expect a brand new model on latest hardware to work faster and more precisely.
The code presented in this text is cross-platform, so readers who don’t have a Raspberry Pi can run it on a Windows, Linux, or OS X computer as well.
Without further ado, let’s see how it really works!
Raspberry Pi
For somebody who could have never heard concerning the Raspberry Pi, let’s make a brief…
