Ultralytics provides YOLO models for real-time computer vision tasks including detection, segmentation, classification, and tracking. It offers a unified CLI and Python API for model training, evaluation, and deployment across image and video data.
Project overview
The framework bundles state-of-the-art YOLO architectures into a unified developer interface with automatic weight downloading and multi-format export capabilities.
Project type
Model Runtime · Image & Vision · Video
Deployment
Refer to project documentation
License
AGPL-3.0
Best for
Developers, AI engineers, and researchers who need a unified Python or CLI interface for training and deploying SOTA YOLO models on image and video data.
Key capabilities
Detect and localize objects within images.
Perform instance segmentation to identify object boundaries at the pixel level.
Perform semantic segmentation to classify each pixel in an image.
Classify the primary subject of an entire image.
Estimate the pose or keypoints of objects within an image.
Track detected objects across sequential frames or video.
Train models on custom datasets using the CLI or Python API.
Export trained or pretrained models to various formats like ONNX for deployment.
Limitations and risks
The AGPL-3.0 license requires open-sourcing derivative works, which may create compliance hurdles for closed-source commercial applications.
Models do not ship offline; they download automatically from the latest Ultralytics release on first use, which requires internet connectivity during setup or initial inference.
Getting started
Installation is straightforward using a simple pip install command with minimal boilerplate, requiring optional coding knowledge to interact via CLI or Python library.
Alternatives and comparisons
Provides a PyTorch model for vision AI tasks including object detection, instance segmentation, and image classification.
A PyTorch implementation of YOLOv3 providing training, validation, inference, and multi-format export for real-time object detection.
README: creates cutting-edge, state-of-the-art (SOTA) YOLO models built on years of foundational research in computer vision and AI. Constantly updated for performance and flexibility, ou…
README: Install the `ultralytics` package, including all requirements, in a Python>=3.8 environment with PyTorch>=1.8.
README: You can use Ultralytics YOLO directly from the Command Line Interface (CLI) with the `yolo` command: