Tag
LingBot Vision, a self-supervised vision backbone family from Ant Group, uses masked boundary modeling to achieve state-of-the-art performance on dense spatial perception tasks, beating the larger DINOv3 model on NYU-Depth v2.