Building a tree map of an area is a labor intensive task, since trees may cover large areas, e.g., a forest, and may not be easily accessible, e.g., in private gardens. However, having a detailed tree map may help predicting seasonal pollen spread in cities, forecasting harvest on the land, etc.
As part of this thesis we would like you to build a tree map of urban arias such as Graz and Vienna from satellite images. You will build a deep neural network, train it on the data sets of 200k indexed public trees in Vienna and test your algorithm on satellite images from different cities and seasons. We are interested in predicting detailed information about the trees, such as their age and species.
Download as PDF