nvidia/visual-changenet

PREVIEW

Visual Changenet detects pixel-level change maps between two images and outputs a semantic change segmentation mask

Input

This demo identifies pixel level changes in two images, reference and test, taken at different point in time. It uses the VisualChangeNet model trained on land satellite imagery dataset to identify changes in land over time. This is a semantic segmentation model that identifies 10 different types of changes in landscape.

Output

Output image for Visual Changenet
Legend
Farmland to desert
Building to desert
Water to desert
Farmland to building
Desert to building
Desert to farmland
Building to farmland
Water to farmland
Desert to water