license: apache-2.0
pipeline_tag: image-segmentation
library_name: pytorch
UAGLNet
UAGLNet is an Uncertainty-Aggregated Global-Local Fusion Network designed for building extraction from remote sensing images. It exploits high-quality global-local visual semantics under the guidance of uncertainty modeling, addressing challenges posed by complex structural variations. The network features a novel cooperative encoder (hybrid CNN and transformer layers), an intermediate cooperative interaction block (CIB), a Global-Local Fusion (GLF) module, and an Uncertainty-Aggregated Decoder (UAD) to enhance segmentation accuracy by explicitly estimating pixel-wise uncertainty.
📄 Paper: "UAGLNet: Uncertainty-Aggregated Global-Local Fusion Network with Cooperative CNN-Transformer for Building Extraction" (arXiv:2512.12941) 💻 Repository: https://github.com/Dstate/UAGLNet
Sample Usage
You can quickly reproduce the main results for various datasets by running Reproduce.py, which will load the pretrained checkpoints from Hugging Face and perform inference.
# To reproduce results on the Inria dataset:
python Reproduce.py -d Inria
# To reproduce results on the Massachusetts dataset:
python Reproduce.py -d Mass
# To reproduce results on the WHU dataset:
python Reproduce.py -d WHU