Answers are grounded in this paper and cite their sources.
Demonstrates effective segmentation outcomes in traditional villages, highlighting conservation needs.
Rapid urbanization in China increasingly threatens traditional villages with heritage value, necessitating effective monitoring tools. However, traditional segmentation methods remain challenging in traditional villages because village scenes contain irregular building layouts, vegetation occlusion, and complex roof boundaries. To address these issues, we construct a UAV dataset covering 12 traditional villages in coastal Quanzhou and propose DINO-Seg, a DINOv3-based segmentation method for landscape analysis. The model combines a frozen DINOv3 encoder with a resampling block that reconstructs multi-resolution features, a fusion upsampling decoder that restores spatial detail, and a lightweight segmentation head for pixel-wise prediction. Results show that DINO-Seg achieves the best performance among the compared baselines, with average precision, recall, F1-score, and IoU of 0.8363, 0.8702, 0.8522, and 0.7445, respectively. The method identifies traditional buildings, new buildings, greenery, and water bodies, and reveals insights regarding building encroachment, greenery imbalance, and water-body distribution, thereby providing an efficient tool for conservation assessment.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: