The first paper on YOLO was presented late 2015 / early 2016, and when the CNN course was being developed, it was absolutely at the forefront of the computer vision pack. I think it is important to understand what was innovative, what problem with then-current practice it solved, and what problems remained. That said, the lead investigator, Joseph Redmon, stopped research in computer vision after YOLO v3 in 2018 ( https://arxiv.org/pdf/1804.02767.pdf ) and the original lineage of YOLO from his lab is no longer the latest. For that, you should probably take a look at the later series of releases, including several from a company named Ultralytics. They call theirs YOLO v8 if I’m not mistaken, even though it deviates significantly from the architecture of versions 1 ~ 4. No anchor boxes, for example. Elsewhere in this forum there is a link I posted to a paper that covers all the YOLO variants. Probably worth a look.
4 Likes