This project is a general-purpose visual foundation model opened sourced by Meta, including a complete pre-trained backbone network, adapters, training and evaluation code. It can be directly applied to various mainstream visual tasks such as classification, retrieval, depth estimation, and object detection without fine-tuning through self-supervised learning (SSL) technology, and is suitable for fields such as autonomous driving, remote sensing, security monitoring, and industrial inspection.