YOLO stands for You Only Look Once. It’s an object detection model used in deep learning use cases. In this article, I will not talk about the history of the previous YOLO versions or the background of how it works. If you are interested in details on how it works, you can check my previous articles about it.

No one in the research community expected a YOLOv4 to be released…

Why? Joseph Redmon, the original author, had announced that he would stop doing computer vision research because of the military and ethical issues…

But someone had continued the hard work… and it was Alexey Bochkovskiy!

YOLO belongs to the family of One-Stage Detectors (You only look once — one-stage detection). One-stage detection (also referred to as one-shot detection) is that you only look at the image once. If we would need to answer what does it mean in fewer sentences it would sound the following:

• It is a sliding window and classification approach, where you look at the image and classify it for every window;
• In a region proposal network, you look at the image in two steps — the first to identify regions where there might be objects, and the next to specify it.

YOLOv4

In this article, we’ll try to understand why the release of YOLOv4 spread through the internet in just a few days. Why it’s called a super-network that can, once again, change the world, same as YOLOv3 did.

Most people in the field today are used to YOLOv3, which already produces excellent results. But now, YOLOv4 has improved again in terms of accuracy (average precision) and speed (FPS)- the two metrics we generally use to qualify an object detection algorithm:

As shown above, YOLOv4 claims to have state-of-the-art accuracy while maintains a high processing frame rate. It achieves an accuracy of 43.5% AP for the MS COCO with an approximately 65 FPS inference speed on Tesla V100. In object detection, high accuracy is not the only holy grail anymore. We want the model to run smoothly in the edge devices. Processing video input in real-time with low-cost hardware becomes important also.

Before diving into the details in this article, I recommend you to read YOLOv4 paper.

#computer-science #deep-learning #data-science #machine-learning #computer-vision #deep learning