The idea of the project work to use the knowledge that you have gained during the course to solve real life problems. Additionally, you are expected to report your work in a similar manner as you do in a scientific publication (with a considerably smaller scope). Together with your group, you are expected to consider a suitable challenge, choose what type of deep learning model would be suitable, make the implementation and report your findings. We will offer you pointers to various datasets that can be used in your project. You are not limited to any particular data, a deep network structure or a computational environment.
Instructions for selecting a topic for the project
The topic your group selects should be such that it is interesting and challenging, but not too challenging so that you can accomplish it all (literature survey, code writing, debugging, experimenting, reporting) in 30 hours per student. It is advisable to select a topic and research problem for which there exist a well-defined evaluation scheme and reference results you can compare against.
The project proposals should be submitted using the tool on this web page.
Notice: If you already have done or are currently doing a similar project in another course, it is not acceptable to report the same project in this course!
You can do your project alone or in a team. The recommended size of a team is two people but I can accept teams with three people if you convince me that the project is challenging enough. You need to declare your team when you submit the project proposal.
Web resources for the project
There are good resources in the web where you can find problems and datasets for your deep learning research. These include but are not limited to: