There are numerous methods for out-of-distribution (OOD) detection and related problems in deep learning, see e.g. [1] for an overview. Many of these however only work well in highly fine-tuned settings and are not well understood in broader context. In this project, you would implement a selection of OOD methods for image classification networks and investigate their relations and performance in various circumstances, such as under varying model accuracy. This could be a broad comparison as well as deeper analysis of a specific type of method.  

You should be comfortable with the implementation of image classification networks and setting up larger experiments. The project will have a significant amount of coding, and independence in that part is appreciated.  

[1] Yang, J., Zhou, K., Li, Y., & Liu, Z. (2021). Generalized out-of-distribution detection: A survey.