Ultralytics YOLO

Ultralytics YOLO is a cutting-edge, open-source computer vision framework built on years of foundational research in AI and deep learning. It provides a family of fast, accurate, and easy-to-use YOLO models that excel at object detection, instance segmentation, image classification, pose estimation, and oriented bounding box detection. The library is installable via pip and supports both a CLI and a Python API, making it accessible to researchers and production engineers alike. It is licensed under AGPL-3.0 for open use, with an Enterprise License available for commercial deployments.

• Object Detection — Run pretrained YOLO26 models on images or video streams with a single yolo predict command or Python call.
• Instance Segmentation — Use YOLO26-seg variants to produce pixel-level masks for each detected object.
• Image Classification — Classify images across 1000 ImageNet classes using YOLO26-cls models pretrained for high top-1 accuracy.
• Pose Estimation — Detect human keypoints with YOLO26-pose models trained on COCO-Pose.
• Oriented Bounding Boxes (OBB) — Detect rotated objects in aerial imagery using YOLO26-obb models trained on DOTAv1.
• Object Tracking — Apply multi-object tracking across video frames; compatible with all detection, segmentation, and pose models.
• CLI & Python API — Install with pip install ultralytics, then use the yolo CLI or import from ultralytics import YOLO in Python.
• Multiple Export Formats — Export models to ONNX, TensorRT, CoreML, TFLite, and more for cross-platform deployment.
• Integrations — Native support for Weights & Biases, Comet ML, Roboflow, Intel OpenVINO, Neural Magic DeepSparse, and the Ultralytics Platform.
• Multi-language Docs — Full documentation available in Chinese, Korean, Japanese, Russian, German, French, Spanish, Portuguese, Turkish, Vietnamese, and Arabic.