Deformity is a serious health issue in aquaculture and can take various forms. Some are difficult for visual detection, and often experts' assessments refer to deformity as a general term without much subtlety. Different deformities can have different physiological and biological causes, and different genetic underpinnings. Advances in image analysis and artificial intelligence can help detect deformities in real time. However, the reliability and sensitivity of such models to different forms of deformities remain challenging, mainly due to the limited availability and granularity in experts' assessments.
This project focuses on developing explainable computer vision models that break down each type of deformity with respect to specific morphological features. An image dataset of pike perch is available, where fish exhibit different forms and levels of deformities. The student will work on an automated image analysis pipeline where fish can be quantified according to their degree of deformity.
To be decided in agreement with the supervisor.
Center for Quantitative Genetics and Genomics
60 ECTS. Experimental theses in which the student is responsible for planning, trial design and collection and analysis of his/her own original data. Can be adjusted based on the student’s preference and discussion with supervisors.
Requirement: Understanding of statistics, genetics and animal breeding.
Programming skills in Python or image analysis are an advantage but not a prerequisite.
