Master's thesis; AI-based Anomaly Detection and Active Learning in Bioacoustics

Master's thesis; Data-Centric AI for Acoustic Discovery: Anomaly Detection and Active Learning in Bioacoustics

Introduction The AukLab-Audio Initiative
The AukLab-Audio project represents a cutting-edge, long-term data collection effort at the Stora Karls¨o Auk Lab, a unique cliff-edge observatory for common guillemots (Uria aalge) in the Baltic Sea. This initiative is generating an unprecedented multi-channel, multi-modal dataset, featuring continuous, synchronized audio recordings from an array of high-quality DPA microphones and camera-integrated microphones, alongside extensive video footage, weather data, and other sensor information. The richness of this dataset opens up unparalleled opportunities for research at the intersection of bioacoustics, machine learning, and marine ecology. We are excited to offer four interconnected Master’s thesis projects designed to explore different facets of this unique dataset. These projects aim to foster a collaborative research environment, where students will contribute distinct expertise towards a unified goal: to deepen our understanding of guillemot vocal behavior, develop novel computational analysis techniques for complex acoustic scenes, and derive new ecological insights. A potential research field trip to the Stora Karlsö research site may be organized to provide students with first-hand experience of the data collection environment and the study species. We invite applications from motivated students for the following project:

Master’s thesis project

• Focus: This project emphasizes data-centric approaches to uncover novel acoustic patterns and efficiently expand our knowledge of guillemot vocalizations. The student will develop methods for anomaly detection in long-term acoustic recordings and implement active learning strategies to guide the annotation process.

• Potential Tasks:

• Designing and evaluating anomaly detection algorithms (e.g., using prototypical networks, autoencoders, or other unsupervised methods) to identify unusual or rare sound events within the vast dataset.
• Developing active learning frameworks that intelligently sample audio segments most likely to contain new or underrepresented vocalizations, thus optimizing human labeling efforts.
• Investigating methods for semi-automatic expansion of acoustic vocabularies by incorporating feedback from anomaly detection and active learning loops.
• Creating tools or workflows that assist human annotators by prioritizing data segments flagged by these intelligent systems. E.g, extend an audio annotation tool with the developed methods.

• Main Supervisor: John Martinsson

Collaborative Environment and Unified Goals
Several master’s thesis projects will run at the same time and are designed to be highly synergistic. Regular joint meetings and knowledge sharing will be encouraged to ensure that developments in one project can inform and accelerate progress in others. The ultimate aim is to collectively advance our understanding of common guillemot ecology through innovative computational analysis of the rich AukLab-Audio dataset, contributing to both ecological science and machine listening methodology.

Who are you?
We are looking for two motivated students with a solid background in Python and machine learning, particularly in areas such as unsupervised learning, active learning, and data-centric AI. You should have strong analytical and programming skills, with experience in machine learning frameworks such as PyTorch or TensorFlow. An interest in applied AI and environmental applications is highly valued. Experience with audio processing, large datasets, or acoustic data is considered an advantage.

What We Offer

• Supervision by an interdisciplinary team of researchers with expertise in machine learning, bioacoustics, and marine ecology.

• Access to the unique, large-scale AukLab-Audio dataset and computational resources.
• An opportunity to contribute to cutting-edge research with potential for high-impact publications.
• A collaborative and stimulating research environment.
• Potential for a research field trip to the Stora Karlsö field station. 
• After a successful project completion, RISE will pay each student SEK 30 000.

Terms

• Location: Gothenburg
• Time: January to June, 2026
• Suitable for a team of two students
• Contact: Olof Mogren, olof.mogren@ri.se and Main Supervisor: John Martinsson

Welcome with your application!
Interested candidates should submit their application at the latest November 30, 2025:

• A curriculum vitae (CV).

• Academic transcripts. <