Embry-Riddle Coastline Dataset (ER-COAST)

The Embry-Riddle Coastline Dataset (ER-Coast) is designed for maritime machine learning research, including sensor fusion. ER-Coast introduces a diverse set of data using multiple sensing modalities. These include three LiDAR, three 8-megapixel color cameras, one 5.4-megapixel high dynamic range camera, two long-wave infrared cameras, and a GPS/IMU inertial navigation system. The captured data includes coastal waterways in the state of Florida during the day and at night. There are 36 separate sequences split across 4 days of collections, totaling over 5 hours of calibrated and timestamped data. A subset of this raw data has been annotated for the tasks of LiDAR semantic segmentation, image semantic segmentation, and image object detection. Intrinsic and extrinsic calibrations are available in the dissertation of https://commons.erau.edu/edt/741/, and the original calibration data has also been provided for users to generate their own calibrations. A paper with a focused discussion of ER-COAST will be released at a later date.