Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Vision Pretraining for Dense Spatial Perception

About

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.

Zelin Fu, Bin Tan, Changjiang Sun, Shaohui Liu, Kecheng Zheng, Yinghao Xu, Xing Zhu, Yujun Shen, Nan Xue• 2026

Related benchmarks

TaskDatasetResultRank
Semantic segmentationADE20K
mIoU53.5
699
Semantic segmentationCityscapes
mIoU79.6
526
Semantic segmentationPASCAL VOC 2012
mIoU87.5
231
Semantic segmentationPascal VOC
mIoU87.4
214
Depth EstimationNYU V2
RMSE0.31
207
Depth EstimationKITTI
RMSE2.552
184
Image ClassificationImageNet-1K
Top-1 Accuracy86.38
60
Semi-supervised Video Object SegmentationDAVIS 2017 (val)
J&F Score70
55
Depth CompletionNYU V2
RMSE0.113
51
Depth EstimationNYU V2
RMSE0.296
22
Showing 10 of 26 rows

Other info

GitHub

Follow for update