AI RESEARCH
dinov3.seg: Open-Vocabulary Semantic Segmentation with DINOv3
arXiv CS.AI
•
ArXi:2603.19531v1 Announce Type: cross Open-Vocabulary Semantic Segmentation (OVSS) assigns pixel-level labels from an open set of text-defined categories, demanding reliable generalization to unseen classes at inference. Although modern vision-language models (VLMs) strong open-vocabulary recognition, their representations learned through global contrastive objectives remain suboptimal for dense prediction, prompting many OVSS methods to depend on limited adaptation or refinement of image-text similarity maps.