The primary objective of this project is to develop innovative visual question answering approaches for imagery collected by heterogeneous sensors to enable machines to understand the semantic meaning of a scene from drones. The emergence of small unmanned aerial vehicles along with inexpensive sensors presents the opportunity to collect millions of images every day with high flexibility and easy maneuverability. Despite the ease of data collection, data analysis of the big datasets collected by various sensors remains a significant barrier for scientists and analysts. While traditional analyses provide some insights into the data, the complexity, scale, and multi-disciplinary nature of the data necessitate advanced, intelligent solutions. Imagine that a pasture manager, looking at images collected over his herd, could ask, “Are there any sick animals in the herd?” or a disaster manager could ask, “How many people have injuries?” or a climatologist, when examining radar images taken from an airplane, could ask “What will the ice thickness be next month?” Quick identification of sick animals would reduce the spread of illness through the herd. Reducing the time spent in disaster response saves human lives, and knowledge about melting of polar ice sheets could help improve plans for evacuation of coastal towns affected by sea level rise. After or during each flight mission, scientists could ask their questions directly to their machines and get a prompt and accurate answer, managing environmental resources, saving human lives, and making intelligent decisions would become faster and more efficient. In this project, students will develop some novel visual question answering techniques for drone images and test it in some real-life applications including drone images collected after natural disasters.