Student Assistant
Data-efficient Training of neural networks for image segmentations (Id 321)
No current offer!
Image segmentation involves the assignment of image pixels to object categories. While architectures for image segmentation are well established, annotation of image data is still a resource-intensive task. To minimize the data volume requirements for training such neural networks, approaches that exploit known symmetry transformations of the target objects have been proposed in recent years. In this work, different approaches of such equivariant networks will be implemented, trained and compared.
Department: Computational Science
Contact: Dr. Steinbach, Peter
Requirements
- python programming & insights into group theory (essential)
- basic understanding of neural networks (especially CNNs or U-nets)
- basic understand of computer vision and image analysis
Conditions
We offer payment of up to 19h hours of work per week. The place of service can be freely chosen - remote work is preferred. There will be supervision of the student (f/m/d) by active scientists (f/m/d).