TreeFusion v1.0
收藏资源简介：
TreeFusion v1.0 is a very-high-resolution urban forest dataset designed for urban tree canopy analysis. The dataset provides two geospatial products for 50 U.S. cities core urban region: Canopy Height Map (CHM) and Vegetation Mask. The products were generated using a multimodal fusion model that integrates NAIP RGB-NIR imagery, Sentinel-1 SAR VV/VH backscatter, normalized Digital Surface Model (nDSM), and rough canopy height priors. The model was trained at large scale using the OpenTree400k benchmark and achieved strong performance with an MAE of 2.09 m and an R² of 0.78. To further evaluate generalization ability, the TreeFusion model was independently validated using 10 selected very-high-resolution NEON ALS sites, achieving an MAE of 2.04 m and an R² of 0.71. Cross-city experiments were also conducted to assess the scalability and transferability of the model across diverse urban environments. If you need data expansion, you can contact us through: taige@vt.edu
