Master thesis – Feature learning for salmon image recognition – Vattenfall R&D
Power Climate Smarter living – that’s our purpose. We are looking for the brightest students to make this happen. Together you will contribute to a fossil free future. You will have a unique position to contribute to our purpose. All we ask from you is to use your Superpowers and share your energy to help us achieve our goal.
The Vattenfall R&D section for Data Science & AI apply advanced analytics to support the mission of fossil free living within one generation. In one of our current projects image recognition is used to identify different types of fish travelling the fish ladder of one of our hydro power plants. Guaranteeing a healthy ecosystem in our rivers is a high priority for Vattenfall, and investigating how new tools can support that goal is key for success.
After gathering extensive image data with high quality, one bottleneck is the labeling of these images with the right features, a time consuming process limited to people experienced in fish anatomy. Finding efficient ways of representing the features of the images in our dataset is key to scaling this technology to additional sites around Sweden.
The goal of this thesis is to develop a representation learning method that extract features from the dataset that can provide similar or increased understanding compared to manually annotated dataset.
Scope of the thesis (30HP)
We would like the student to:
1. Explore tools and technologies for feature/representation learning
2. Benchmark results with the already in use image recognition model in the Vattenfall environment
3. Provide recommendations for improvements in current work
What else do I need to know?
• This master thesis is done with Vattenfall R&D together with the research program FiDiMo.
• The location for this work is flexible, but the team is mostly situated at Vattenfall R&D in Solna, Stockholm
You are a computer science or engineering student with knowledge of or interest in feature/representation learning and different machine learning models used for image recognition. You have a keen interest in sustainability and a drive to share your knowledge by developing solutions that can be used broadly with limited analytics knowledge.
Swedish is not required but we appreciate if you intend on learning it during your thesis.
CV, cover letter, and grades are to be submitted in the application.
The interview process will be ongoing, as soon as relevant applicants are identified.
Deadline for application: 2020-12-06.
Do not hesitate to contact us for any questions:
Dela detta jobb