Area estimation and object recognition are common tasks in a wide range of topics, from urban planning and environmental studies to auto-driving. Semantic segmentation is a computer vision task to assign labels to each pixel in an image, and is broadly applied in auto-driving and video & scene understanding. This project aims to develop an approach to area estimation based on semantic segmentation, which could surpass manual work in terms of time, efficiency and accuracy. Combined with object detection and recognition, this could be applied on USVs (unmanned surface vehicles) to assist navigation and obstacle avoidance, and on UAVs (unmanned aerial vehicles) for wider area estimation. Testing will be performed on USVs and drones for real-life applications.