Shulou(Shulou.com)06/01 Report--

In this issue, the editor will bring you about the neural network how to detect the target YOLO in full field of vision in real time. The article is rich in content and analyzes and describes for you from a professional point of view. I hope you can get something after reading this article.

YOLO's full name is You only look once (you only need to take a look at it), which is a real-time target detection system for cameras. It can distinguish 6000 objects and can process video at a frame rate of 40-90FPS on a Titan X card.

No matter how fast 007 runs, it can remember at a glance:

YOLO, written by Joseph Redmon of the University of Washington (Paul Allen's alma mater) and Ali Farhadi of Paul Allen AI Institute, is currently open source, written in C and CUDA, and has trained parameters for you to download.

YOLO is different from previous target detection systems in principle. In the past, people only reused the models of classifiers and locators to target detection to monitor multiple locations and areas in the camera field of vision, and the area with the highest score was considered to be the target.

YOLO's neural network monitors the entire field of view of the camera, as shown in the following image, which divides the image of the entire field of view into 13 × 13 grid cells:

Each grid cell is responsible for predicting five target boxes, and using the target box to describe the objects detected by the neural network:

However, the confidence value output by YOLO is not for the target it wants to identify, but for the fit of the shape of the target box. The higher the confidence, the thicker the target box:

After the target box is determined, the grid cells will predict the classification of the target accordingly. Taking the PASCAL VOC image dataset as an example, YOLOh can easily identify 20 different targets: bicycles, boats, cars, cats, dogs, people.

Different from the old classifier-based system, YOLO can detect targets in real time by running only one neural network, which is 1000 times faster than the R-CNN system which needs to run thousands of neural networks to detect targets.

This is how the neural network shared by Xiaobian can detect the target YOLO in real time. If you happen to have similar doubts, you might as well refer to the above analysis to understand. If you want to know more about it, you are welcome to follow the industry information channel.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.