Round Corner
Department of Computer and Information Science


Explainable AI for Maritime Situational Awareness (AI Lab pitch)

Problem Description: The main goal of this research project is to explore the feasibility and usefulness of using computer-graphic simulators for development and validation of explainable AI techniques in maritime scenarios. The work falls within the area of computer vision as an enabling  technology for autonomous ships. Deep networks have shown astonishing performance in image recognition and classification tasks. However, effective training requires huge amounts of images (with corresponding labels), whose acquisition (and preparation) is very expensive. As alternative solution, Kongsberg Digital (in collaboration with NTNU) has explored the use of synthetic computer-generated data in combination with domain adaptation techniques. Results have been promising in aquaculture and maritime domains, where satisfactory performance have been achieved with limited effort on datacollection and preparation. Building on that, Kongsberg Digital wants to investigate the relationships among synthetic-vs-real images in the framework of explainable AI, to assess if computer-graphic simulators and domain adaptation techniques can facilitate and/or improve the design of decision-support systems providing predictions coupled with explanations. 

Problem Motivation
Even with excellent performance, a black-box approach is not suitable for industries with safety-critical and/or economically-impactful issues. The lack of explanations to be coupled with predictions may prevent machine-learning models from being fully exploited in the maritime industry. Trustworthy and understandable predictions are necessary for including machine-learning techniques in decision-support systems (both in semi-autonomous and fully-autonomous scenarios) and would greatly facilitate the adaptation of such technology in the maritime domain. 

Data Availability
Two Kongsberg Digital’s simulator will be considered as sources of data:

  • Cogs, developed by the 3D team;
  • K-Sim, developed by the maritime simulator team.

The student will be given access to the simulators and will be exposed to the tools for data manipulation considered in the previous works. Datasets well suited for task can be generated as part of the work and will not be subject to any restriction/protection.

Description of work:

Theoretical work includes:

  • Literature study of Mask R-CNNs for image segmentation;
  • Literature study of domain adaptation in image classification;
  • Literature study of explainable AI techniques in computer vision.

Practical work includes:

  • Synthetic image generation using Kongsberg Digital simulators (Cogs and/or K-Sim);
  • Implementation of training procedures for image segmentation and domain adaptation techniques (Python code from previous projects is available);
  • Development of a framework for explainable AI for maritime situational awareness;
  • Performance analysis and comparison in a few selected cases of interest.

Contacts outside the AI Lab:

  • Pierluigi Salvo Rossi (
  • Thorvald Grindstad (
  • Christopher Dyken ( Lekkas ( – NTNU Dept. Eng. Cybernetics
  • Theoharis Theoharis ( – NTNU Dept. Computer Science


Other information:

This is a project in collaboration with an external partner. If you choose this project, then I will serve as the responsible from NTNUs side, but the actual work will also be in tight collaboration with personell from the external partner as listed above.

If you consider picking a project with me as the supervisor, then please look at




Helge Langseth Helge Langseth
310 IT-bygget
735 96488 
NTNU logo