DINOv2 alternatives
DINOv2 is a family of foundation models designed to generate universal visual features for both image-level and pixel-level tasks. These models are trained without labels, enabling robust and versatile features that can be used directly in various computer vision applications.