Just wanted to put out a quick note that YOLOv7 is here. Things sometimes move lightning fast in the state of the art models for object detection. We just wrote about YOLOv6 last week, and were doing a thorough evaluation, and now we have a more recent model to evaluate. The YOLOv7 paper just came out on arxiv.org, and it is by the same authors who worked on YOLOv4.

Is YOLOv7 is the right model for you?

Remember that although the paper claims better performance compared to previous models (YOLOv6 was not included in the comparison) on the COCO dataset, you need to validate if these claims really work in your dataset. We will aim to do the same on the Anki Vector dataset, with the help of a modified notebook that our friends at Roboflow put up. Roboflow also has a blog article on YOLOv7.

Please keep watching this space as we evaluate these state of the art models for object detection. If you are not yet a subscriber, please join us for free.