GPT-4: Meta AI’s DINOv2 is a groundbreaking self-supervised learning method for training computer vision models, achieving performance on par with or better than the gold standard. The model can learn from any image database without requiring human annotations, making it highly adaptable for various computer vision tasks. With 142 million photos in its dataset, DINOv2 can produce general-purpose visual features that perform well across domains without fine-tuning, simplifying the use of images in AI systems.
Read more at MarkTechPost…
