Dry Harbour Image Dataset for Semantic Segmentation

This dataset was used to produce the results in the paper "Integrated Design-Sense-Plan Architecture for Autonomous Geometric-Semantic Mapping with UAVs", Frontiers in Robotics and AI, 2022.

The dataset consists of training data for a semantic segmentation network targeted at robotic inspection of dry dock shipyard environments.

The data was generated by using a combination of Unreal Engine 4 and AirSim: to create a realistic virtual dockyard environment and to extract images and corresponding labels (Sky, Harbor wall, Floor, Ship, Unknown).