YOLOv5

Ultralytics YOLOv5 provides a high-speed, production-ready object detection framework built on PyTorch for seamless deployment.

YOLOv5 by Ultralytics remains a standard for real-time computer vision applications. It features five distinct model scales (Nano, Small, Medium, Large, and XLarge) to optimize for specific hardware constraints. The YOLOv5s variant clocks in at 2.0ms inference time on an NVIDIA Tesla V100, delivering the responsiveness required for live video feeds. Engineers deploy these models for practical use cases: identifying manufacturing defects, managing retail inventory, and securing autonomous perimeter monitoring. The ecosystem supports direct exports to multiple formats (ONNX, OpenVINO, and CoreML) to ensure compatibility across cloud and edge environments.

https://github.com/ultralytics/yolov5