Internship assignment: "Deep-learning based separation of ship-radiated underwater noise"
Publication date19 Feb 2026
FieldResearch & Development
LocationWageningen
Employment40 hours
For MARIN Academy we are looking for a student for the following internship:
Deep-learning based separation of ship-radiated underwater noise
Internship
Accurate ship detection, recognition and location from underwater noise observations in the North Sea is important in environmental monitoring and maritime defense. The North Sea is one of the busiest maritime areas in the world making it difficult to obtain single-source datasets for machine learning from hydrophone recordings. The presence of multiple sound sources near the measurement location poses the challenge to disentangle these sources for further analysis. Successful separation of noise radiated by different ships using deep-learning is the main goal of the project. Previous research in speech separation has shown that the separation of audio can be done successfully. However, the amount of research for ship source separation is still limited. An example of such a deep-learning architecture is given in Figure 1.
Two phases of the project are defined:
(1) Find a Machine Learning method to automatically detect how many ships are present within the recording using synthetically generated data.
(2) Explore how (the possibilities to apply) these techniques are able to translate to a real world scenario derived from the North Sea. (to combined noise measurements with AIS datasets.)
(3) If possible, Develop and test deep-learning models for separation of underwater noise on public benchmark datasets such as DeepShip and ShipsEar.
The successful candidate develops (novel) deep learning methods able to detect the number of ships recorded and maybe separate ship-radiated noise and evaluates (demonstrates) these techniques on data from hydrophone recordings in the North Sea.
Duration
The internship is for the duration of 5 months. The start date of the internship period can be determined in consultation with the supervisors.
The profile
Department and supervisor
This internship will take place at the department R&D of MARIN. The supervisors for this internship are Hilde Hummel (PhD candidate, VU) and Thomas Scholcz (Senior researcher, MARIN).
You can apply for this internship by using the APPLY button.
Deep-learning based separation of ship-radiated underwater noise
Internship
Accurate ship detection, recognition and location from underwater noise observations in the North Sea is important in environmental monitoring and maritime defense. The North Sea is one of the busiest maritime areas in the world making it difficult to obtain single-source datasets for machine learning from hydrophone recordings. The presence of multiple sound sources near the measurement location poses the challenge to disentangle these sources for further analysis. Successful separation of noise radiated by different ships using deep-learning is the main goal of the project. Previous research in speech separation has shown that the separation of audio can be done successfully. However, the amount of research for ship source separation is still limited. An example of such a deep-learning architecture is given in Figure 1.
Two phases of the project are defined:
(1) Find a Machine Learning method to automatically detect how many ships are present within the recording using synthetically generated data.
(2) Explore how (the possibilities to apply) these techniques are able to translate to a real world scenario derived from the North Sea. (to combined noise measurements with AIS datasets.)
(3) If possible, Develop and test deep-learning models for separation of underwater noise on public benchmark datasets such as DeepShip and ShipsEar.
The successful candidate develops (novel) deep learning methods able to detect the number of ships recorded and maybe separate ship-radiated noise and evaluates (demonstrates) these techniques on data from hydrophone recordings in the North Sea.
Duration
The internship is for the duration of 5 months. The start date of the internship period can be determined in consultation with the supervisors.
The profile
- Master student in Computer Science / Artificial Intelligence, or other applied science.
- Interest in acoustics.
Department and supervisor
This internship will take place at the department R&D of MARIN. The supervisors for this internship are Hilde Hummel (PhD candidate, VU) and Thomas Scholcz (Senior researcher, MARIN).
You can apply for this internship by using the APPLY button.
Contact
HRM