Третья итерация знаменитой архитектуры компьютерного зрения от Meta, работающая на принципах самообучения без ручной разметки данных. DINOv3 устанавливает новые стандарты в классификации и сегментации изображений, легко масштабируясь до гигантских датасетов и сложнейших визуальных задач.
Self-supervised learning holds the promise of eliminating the need for manual data annotation, enabling models to scale effortlessly to massive datasets and larger architectures. This technical report introduces DINOv3, a major milestone toward realizing this vision by leveraging simple yet effective strategies. First, we leverage the benefit of scaling both dataset and model size by careful data preparation, design, and optimization. Second, we introduce a new method called Gram anchoring, which effectively addresses the known yet unsolved issue of dense feature maps degrading during long training schedules. Finally, we apply post-hoc strategies that further enhance our models' flexibility with respect to resolution, model size, and alignment with text.