[CVPR 2019, Oral] "Collaborative Global-Local Networks for Memory-Efficient Segmentation of Ultra-High Resolution Images" by Wuyang Chen*, Ziyu Jiang*, Zhangyang Wang, Kexin Cui, and Xiaoning Qian
-
Updated
Dec 29, 2021 - Python
[CVPR 2019, Oral] "Collaborative Global-Local Networks for Memory-Efficient Segmentation of Ultra-High Resolution Images" by Wuyang Chen*, Ziyu Jiang*, Zhangyang Wang, Kexin Cui, and Xiaoning Qian
DeepGlobe Land Cover Classification Challenge遥感影像语义分割
U-Net road segmentation pretrained on DeepGlobe and fine-tuned on UAVid (IoU 0.7168)
Lightweight U-Net for road segmentation from satellite images using the DeepGlobe dataset.
Semantic segmentation of high-resolution satellite imagery on the DeepGlobe Land Cover dataset, comparing U-Net, DeepLabV3+, and DenseASPP under a controlled multi-scale context modeling framework.
Semantic segmentation of satellite and aerial imagery for land cover classification using U-Net + ResNet-50. Trained on DeepGlobe (7 classes, mIoU 0.637) and LandCover.ai (4 classes, mIoU 0.798). Includes interactive Gradio demo with confidence-based OOD detection and class distribution visualization.
Satellite road extraction pipeline: SegFormer-B2 semantic segmentation on DeepGlobe imagery (0.68 IoU) → morphological post-processing → A* pathfinding on predicted road networks. Includes a Gradio app and Docker image.
To associate your repository with the deepglobe topic, visit your repo's landing page and select "manage topics."