Meta has released DINOv2, the first method to use self-supervised AI learning to generate outcomes on par with or better than industry norms.
The company has made significant progress by updating its visual recognition model, DINO. It now uses self-supervised learning to identify items in images and video frames, instead of relying on human interpretation for accuracy.
Meta will be able to create more accurate representations of DINOv2 and can identify individual elements within a scene and has improved contextual awareness.
Meta is optimistic that the team behind DINOv2 has plans to incorporate this model into more complex and big AI systems.