YOLO (Ultralytics)

Ultralytics YOLO is a Python computer vision library for developers building applications that interpret images and video. Models can run on hardware you control for tasks such as visual inspection, inventory monitoring and robotics perception.

• Image analysis: Detect objects, create instance or semantic masks, classify images, estimate body keypoints and depth, or locate rotated objects.
• Video tracking: Follow objects across frames using detection, segmentation, pose or oriented detection models.
• Custom training: Use pretrained models or train and validate models against your own datasets through Python or the command line.
• Device deployment: Export models to ONNX, TensorRT, OpenVINO, CoreML and LiteRT for use outside the Python library.

Local inference supports CPUs and NVIDIA GPUs. Edge deployment targets include NVIDIA Jetson, Raspberry Pi and mobile phones. Integrations include Weights & Biases, Comet ML and Roboflow for experiment tracking and dataset work. Pretrained models download on first use.

The library is open source under AGPL-3.0, with separate enterprise licensing available. The hosted Ultralytics Platform provides browser-based annotation with SAM assistance, dataset review and versioning, cloud GPU training, and managed prediction endpoints with monitoring. Those hosted functions run on external infrastructure; exported models can run locally.