Review of recent techniques used for Aerial Image Segmentation

The paper deals with aerial image segmentation to make a pixel level classification of buildings(masks) from satellite images. Such techniques are useful for identifying land encroachments at remote places. Existing techniques require a large amount of data to train the models which predict masks. U-Net based models are essentially used for medical image segmentation but this technique can be extended to aerial image segmentation as well. In addition to that several other techniques have come up which require less data and therefore it becomes necessary to discuss the different techniques in existence.