Přejít k obsahu
Cílená práce

Hav-Inno: Using Artificial Intelligence to Identify and Quantify Marine Life from Underwater Videos

Projekt

Henrik Dams Allé, 2800 Kongens Lyngby (Denmark), Vejlsøvej 39, 8600 Silkeborg (Denmark)

Zveřejněno dne 4. listopadu 2026

  • Typ smlouvy

    Projekt

  • Polohy

    Henrik Dams Allé, 2800 Kongens Lyngby (Denmark), Vejlsøvej 39, 8600 Silkeborg (Denmark)

  • Datum začátku

    Prosinec 2026

  • Plat

    Informace nebyly poskytnuty

  • Práce z domova

    částečné

Student projects, bachelor and master's thesis opportunities at DTU Aqua and DTU Compute

Aim

The HavInno project brings together DTU Aqua and DTU Compute to develop AI-based methods for analysing underwater video footage and identifying and quantifying marine life.

The project aims to develop a computer vision model that can automatically detect fish and macro-epifauna and identify them to species level.

By combining marine biology and AI, the Hav-Inno project will make marine biodiversity monitoring more efficient, standardised, and less dependent on time-consuming manual video analysis. Students can cover either the biological aspects of the project at DTU Aqua or the advanced development of AI models at DTU Compute, depending on their interests and background.

The Hav-Inno project offers projects to students from Danish and international universities.

Background

Underwater video footage is an important tool for monitoring marine biodiversity. DTU Aqua has access to a large and growing collection of underwater recordings, documenting marine life across different seabed habitats.

Traditionally, these recordings have been analysed manually. Observers review video frames, identify the organisms present and estimate their maximum observed abundance (MaxN) for each species. However, analysing thousands of video sequences manually represents a substantial workload. Analysing footage by the human eye is not only time consuming and requires substantial training, but it is also prone to human errors. As new recordings continue to be collected, there is a growing need for standardised methods that can support or partially automate this process.

The Hav-Inno project addresses this challenge by combining DTU Aqua's biological expertise and video data with DTU Compute's expertise in artificial intelligence and computer vision. The long-term ambition is to develop a tool that can assist researchers in analysing underwater footage, reduce manual workload, provide standardised quantitative underwater data, and support large-scale marine biodiversity assessments.

Project opportunities

Students can choose between two distinct project tracks within the Hav-Inno project. The scope, academic requirements and expected outputs differ between the two tracks.

1. DTU Aqua – Marine biology, biodiversity and underwater video analysis Who is this for?

Students from a range of academic backgrounds who are interested in marine biology, marine biodiversity, ecology, scientific communication, or data analysis.

No specific technical prerequisites are required. Students do not need previous experience with AI or programming to participate in this track.

Project content

At DTU Aqua, students can contribute to the Hav-Inno project through projects related to underwater video footage, marine species identification and biodiversity documentation. The specific project will be developed in collaboration with the student and tailored to their academic background, interests, study requirements, and career goals.

DTU Aqua opportunities include:

  • Contributing to fieldwork, deploying cameras and collecting video recordings (to get an impression of the fieldwork, explore videos from similar studies; video1)
  • Analysing underwater video footage, identifying and quantifying fish and other marine organisms.
  • Investigating fish abundance, species composition related to marine biodiversity.
  • Comparing manual observations with AI-generated detections or classifications.
  • Exploring how automated video analysis can support marine ecological research.
  • Communicating scientific findings to wider audiences (e.g. article or film)

These are examples rather than an exhaustive list of tasks.

Project format and duration

There is no predefined project format, duration, or ECTS requirement. These will be agreed upon between the student and the DTU Aqua supervisor, depending on the student's interests, study programme and academic background.

Possible course outputs can include the development of a popular science article (such as this one), an educational or scientific video (like this), a scientific article (as this one), a report, or another suitable project deliverable. Projects may also be designed as a bachelor's thesis, master's thesis or another type of student collaboration.

What you will gain

Students will gain experience working with real research data and contribute to scientific efforts to understand and document marine biodiversity. Depending on the project, students may develop skills in fieldwork, species identification, data analysis, scientific communication and the use of underwater video footage in marine research. Students can also expand their professional network by joining the Hav-Inno project, as well as work with local NGOs (e.g., Køge Bugt Stenrev or WWF).

2. DTU Compute – AI, computer vision and deep learning Who is this for?

This track is specifically intended for a master's student planning a thesis project with a strong focus on deep learning and computer vision.

Required prerequisite: Advanced/Introduction to Deep Learning in Computer Vision.

Thesis title: Training Fish Detection and Classification Models from VLM-Derived Pseudo-Labels

Project description

Marine biodiversity documentation relies on video surveillance, yet analysing these large datasets remains constrained by the manual effort required to annotate these videos.

Recent advances in foundational AI models offer a compelling solution by automating feature extraction and open-vocabulary object recognition across diverse visual domains. This project investigates how Vision-Language Models (VLMs) can be used to generate pseudo-labels for underwater fish detection and classification.

By extracting zero-shot annotations from VLMs, the project aims to distill knowledge into specialized detection/classification models without relying on extensive manual labeling.

Key research challenges include filtering noisy pseudo-labels and mitigating underwater visual artifacts such as turbidity and occlusion.

The work will contribute to the development of AI methods that can support the automated detection and identification of fish in underwater video footage.

What you will gain

The student will gain experience with deep learning, computer vision and Vision-Language Models, while working on a practical AI research challenge at the intersection of machine learning and marine biodiversity documentation.

This track is suitable for students who want to explore research in computer vision and develop their technical expertise through a master's thesis.

Duration and ECTS

The duration, scope and ECTS credits depend on the selected project track.

  • DTU Aqua: Flexible and agreed individually with the supervisor, depending on the student's academic programme and interests.
  • DTU Compute: A master's thesis project, with scope and duration agreed upon with the supervisors in accordance with the relevant study programme.

Supervisors and contact

DTU Aqua – Marine biology and video analysis

Supervisor: Jon C. Svendsen

Email: < e-mail smazán z bezpečnostních důvodů >

Phone: 00 45 93 51 16 63

Co-supervisors: Lars Emil Juel Andersen

Email: < e-mail smazán z bezpečnostních důvodů >

Co-supervisors: Yasmin Sassa Victor

Email: < e-mail smazán z bezpečnostních důvodů >

DTU Compute – AI and computer vision

Supervisor: Juan Miguel Valverde

Email: < e-mail smazán z bezpečnostních důvodů >

Co-supervisors: Damiano Orlandi

Email: < e-mail smazán z bezpečnostních důvodů >

Interested?

When applying, please indicate which track you are interested in: DTU Aqua, DTU Compute, or both.

For the DTU Aqua track, please briefly describe your academic background, interests, and the type of project you would like to pursue.

For the DTU Compute track, please include your relevant background in deep learning and computer vision and confirm that you are interested in undertaking a master's thesis.

Importantly, please include preferred start and end dates of your project (e.g. January – August 2027)

You are welcome to include links to relevant coursework, projects or other experience.

Uzávěrka pro zasílání přihlášek

1. ledna 2031

Úroveň vzdělání

Žádná konkrétní úroveň

Kategorie pracovních pozic

Životní prostředí a udržitelný rozvoj

Více informací o společnosti

DTU - Danmarks Tekniske Universitet (Network Profile)

At DTU, you get great job flexibility, room for development, and freedom with responsibility