Vector-to-Image Converted Building Footprints for Building Change Detection

We present a challenging change detection problem in vector building footprints, which is critical when detected changes are used to update existing vector databases. We formulate our change detection into an image recognition problem, where deep learning approaches can be used. And we explore a more explainable multi-stage workflow of deep models.1．<b>Python environment</b>: requirements.txt2．<b>Step-by-step instructions on reproducing the results</b><b>:</b> Instructions_on_reproducing.docx3．<b>Code</b><b>:</b> Code_Feature-Fusion-CNN.zip4．<b>Data</b><b>:</b> (1) bounding-box.zip, (2) fixed-window.zip, (3) displacement.zip, (4) discrepancy.zip,(5) window_size_10.zip, (6) window_size_20.zip, (7) window_size_30.zip, (8) window_size_40.zip<br><br>