Year: 2025
Type: article
Abstract: : The YOLOv10 explores a cutting-edge advancement in real-time object detection, widely used in robotics, autonomous vehicles, and surveillance for its enhanced speed and accuracy. YOLOv10 builds on earlier versions by integrating improved convolutional layers, anchor boxes, and transformer-based modules, enabling more efficient object identification in a single neural network run, ideal for time- sensitive applications. The research examines advanced training techniques such as refined data augmentation, optimization, and novel loss functions, with tests on datasets like COCO and PASCAL VOC showing superior accuracy in complex environments, including extreme occlusions and dynamic lighting. Key findings highlight YOLOv10's improved detection accuracy, faster processing, and robustness, as well as its scalability for diverse hardware configurations, making it crucial for intelligent systems in dynamic real-world contexts. These have some Limitations ,Those are The number of objects YOLOv10 can find in an image depends on things like how complicated the scene is, the size of the objects, and if they are blocking each other. However, YOLOv10 is very efficient and can usually detect many objects— sometimes dozens or even hundreds—at once, as long as they are in the categories it has been trained to recognize.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4408748030