Develop a sidescan sonar based detector for the Beckholmen dry dock environment.
Scan of the relevant dry dock at Beckholmen

- Currently, SAM is limited to very basic automatic operations or teleoperation at the surface.
- Limited ground truth is available
- Operations will be at the surface, so the altitude will be constant.
Experiment with deep learning methods to classify the sss returns. I'm currently unsure about the details, but they will be filled in here as they become available.
- Should I use the raw returns on maybe the down range gradients that were used for the buoy and rope detectors?
- Process channels together or separately?
- What am I not considering?
