We are seeking master's students to contribute to a research project focused on developing advanced methods for detecting fish schools using multi-frequency marine echosounder data. The project involves applying pretrained Vision Transformers (ViT) to enhance the differentiation of semantic regions in underwater images, particularly in scenarios where labeled data is limited.
The research will explore novel approaches for improving image analysis, including the development of new graph structures and the application of information-theoretic learning. The objective is to achieve accurate segmentation of objects in complex underwater environments, with an emphasis on both coarse and fine-grained image partitioning.
This project offers an opportunity to engage with advanced methodologies in computer vision, with the potential for contributing to conference presentations. It is well-suited for students interested in applying AI techniques to environmental and marine science research.
References
[1] Sriram Ravindran, Debraj Basu; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 723-733
[2] Fang, Pengfei, Mehrtash Harandi, and Lars Petersson. "Kernel methods in hyperbolic spaces." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2021.