YOLO

YOLO is an object detection algorithm that simultaneously predicts the location and class of all objects in an image through a single pass of a neural network. YOLO, which holds a prominent place in the fields of artificial intelligence and deep learning, is particularly preferred in video processing and autonomous systems due to its real-time analysis capability. YOLO’s most distinctive feature is its ability to detect objects in a single step, delivering efficient results in terms of both speed and accuracy.

^{Computer Vision Tasks Supported by YOLO11.}Ultralytics/Figure 3⁾

The YOLO algorithm was first developed in 2016 by Joseph Redmon and Ali Farhadi. Unlike traditional region-based approaches, this algorithm aims to detect objects by processing the entire image through a single Convolutional Neural Network (CNN).

With successive updated versions, the balance between accuracy and speed has been progressively improved:

• YOLOv1: The initial version established the fundamental working principle.
• YOLOv2 (YOLO9000): Enhanced support for more object classes and improved bounding box prediction.
• YOLOv3: Achieved better performance on small objects through multi-scale prediction.
• YOLOv4: A balanced architecture developed with community contributions, offering strong performance in both speed and accuracy.
• YOLOv5–v8: Modern, modular, and mobile-compatible versions based on PyTorch.

YOLO divides the input image into an S×S grid. Each cell determines whether an object is present and, if so, predicts the object’s class and location coordinates. Each prediction includes the class probability, x and y coordinates of the bounding box center, and width and height values. The results are evaluated using the IoU (Intersection over Union) metric to determine accurate detection boxes.

This architecture enables YOLO to produce high-accuracy results with significantly fewer computations compared to traditional methods. This characteristic makes it especially suitable for applications requiring real-time performance, such as video streams.

^{YOLO divides the image into grid cells and predicts the presence of objects in each cell. (}Redmon et al. 2016⁾

The YOLO algorithm is effectively used across a wide range of industries:

Cameras mounted on vehicles instantly detect traffic signs, pedestrians, and other vehicles to enhance driving safety is provided.

In closed-circuit television systems, real-time monitoring enables human detection, behavior analysis, and threat identification is done.

In production lines, YOLO is applied in quality control, product classification, and sorting systems is used.

In imaging systems, such as detecting abnormal regions in lung X-rays, YOLO supports medical applications available.

In a study conducted by Kaya and Akgül, the YOLOv2, YOLOv3, and YOLOv4 algorithms were comparatively tested on a traffic sign dataset. The following table summarizes their performance metrics: