Roboflow Universe

About Roboflow Universe

Roboflow Universe is an open source repository of computer vision datasets, models, and APIs, hosting over 1 billion images across more than 1 million datasets and 250,000 fine-tuned models. Content is organized by task type, including object detection, instance segmentation, classification, keypoint detection, semantic segmentation, and multimodal, and by application category such as manufacturing, agriculture, construction, logistics, sports, self driving, gaming, and documents.

The platform hosts individual community-contributed projects, each listing its image count, number of trained models, and usage statistics, covering subjects like license plate recognition, PPE and safety vest detection, fire and smoke segmentation, face detection, and sports analytics. It also provides pre-trained YOLO-OCR and YOLO-Face models for text/document reading and face detection respectively, and includes a chat-based agent feature for building computer vision applications from natural-language descriptions.

It is intended for developers and researchers building computer vision applications who need existing labeled datasets or pre-trained models, or who want to publish and share their own vision AI projects within a community.

Key features

• Browse over 1 billion open source computer vision images across 1M+ datasets
• Access 250,000+ fine-tuned models contributed by the community
• Use pre-trained YOLO-OCR models to read text, numbers, receipts, and invoices
• Use pre-trained YOLO-Face models to detect faces in images and video
• Roboflow Agent lets you describe a vision task in chat to help build it

Frequently asked questions