This repository contains several building extraction models which I practiced over time. The task was originally my bachelor thesis project building extraction with convolutional neural network, when I used pixel based segmentation only, and non-efficient seen from now. Since then there are many more effective deep learning algorithms, I tested several of which and collect them here. I'll also try to keep updating it.
-
Pixel based segmentation: simple CNN & Residual Network.
-
Fully convolutional neural network, based on the paper Fully Convolutional Networks for Semantic Segmentation.
-
U-Net, based on the paper U-Net: Convolutional Networks for Biomedical Image Segmentation.
-
Modified U-Net, based on the paper BRRNet: A Fully Convolutional Neural Network for Automatic Building Extraction From High-Resolution Remote Sensing Images.