Arbetsbeskrivning

We want to be more! The research institute Swerim conducts needs-based industrial research and development concerning metals and their route from raw material to finished product. Swerim has 200 co-workers in two locations in Sweden - Luleå and Stockholm. Our vision is a fossil-free and circular industry. Project Description:  This project explores how self-supervised learning can improve defect detection in manufacturing environments where labeled defect data is scarce. The focus is on developing self-supervised pre-training strategies that can learn robust visual representations from unlabeled manufacturing images before fine-tuning on limited labeled defect data. The project will investigate various self-supervised methods such as masked image modeling, or self-distillation to extract meaningful features that transfer well to defect classification tasks. The goal is to investigate whether self-supervised pre-training can significantly improve defect detection accuracy compared to training directly on limited labeled data, making it a viable approach for real-world manufacturing quality control systems.   Qualifications: We are looking for a curious and collaborative Master of Science student with a passion for problem solving and data analysis.   Project time The project is intended for a master thesis (30hp). The start date is January 2026 or can be mutually decided through negotiations.   Further information This project is intended to be performed at Swerim in Stockholm or Luleå. Swerim rewards the student with 50 000 SEK for an approved master thesis (30hp). Contacts For further information please contact: [email protected] Application Apply by using the application function below. The application can be written in English or Swedish. The latest date for application is 30th of November. You will receive confirmation that Swerim has received your application. Please note that we fill the position as soon as we find a suitable applicant, which means we can fill the position before the deadline