Vision Pretraining for Dense Spatial Perception
2026-07-06 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors focus on improving how machines see space by teaching them to understand edges and boundaries in images, which are important for recognizing shapes and depth. They introduce a method called masked boundary modeling, where the model learns to detect detailed edges and uses that knowledge to better understand full images. Using this method, they created LingBot-Vision, which performs well on various visual tasks and helps improve depth estimation, important for machines that interact with the physical world. Their work shows that focusing on boundaries is a useful way to help computers learn about space in images.
dense spatial perceptionvisual foundation modelsself-supervised learningmasked boundary modelingsub-pixel boundary representationsdepth estimationembodied artificial intelligencevisual token learningDINOv3depth completion
Authors
Zelin Fu, Bin Tan, Changjiang Sun, Shaohui Liu, Kecheng Zheng, Yinghao Xu, Xing Zhu, Yujun Shen, Nan Xue
Abstract
Dense spatial perception is essential for physical intelligence, where visual systems are expected to recover structured, metric, and actionable representations from pixel observations. Modern visual foundation models tend to prioritize semantic invariance, often at the expense of detailed spatial understanding. In this work, we study vision pretraining through a boundary-centric lens, motivated by the premise that boundaries and shape discontinuities offer essential cues for perceiving geometric properties. Concretely, we propose masked boundary modeling, a self-supervised paradigm that dynamically learns sub-pixel boundary representations and subsequently leverages the discovered boundary-bearing tokens as masked targets to facilitate dense visual token learning. By scaling this framework, we develop LingBot-Vision and demonstrate its efficacy across a diverse set of downstream vision tasks with DINOv3 as a strong baseline. Remarkably, LingBot-Vision drives the progression from LingBot-Depth 1.0 to LingBot-Depth 2.0 for depth completion, and thereby yields enhanced depth estimation, a key pillar for embodied artificial intelligence. Our findings reveal that boundary modeling goes beyond simple line segments and instead serves as a scalable pretraining principle for learning spatially structured visual representations.