Introduction
In a previous article, we discussed how to evaluate and choose the best Machine Learning (ML) model tailored to your use case. Specifically, we evaluated YOLOv5 vs Scaled YOLOv4, and came to the conclusion that for our specific task of making a Vector robot detect another Vector, Scaled YOLOv4 compared better.
New beasts in town!!!
In the last few weeks, two new versions of YOLO have popped up. We wrote about YOLOv7 and YOLOv6. This article presents an evaluation of YOLOv5 vs YOLOv6 vs YOLOv7 in the context of our specific task: helping Anki Vector detect another Vector in its camera feed.
Evaluating a ML Model
It is important to first lay down the ground rules for doing a thorough evaluation of a Machine Learning apparatus. The ground rules are important to ensure that the comparison made is indeed apples-to-apples, and we are not biased in any way. We have discussed the key concepts of doing an apples-to-apples evaluation before in a previous article. A quick summary is tha…