AAU logo

CMI News

CMI News

PhD project: Models for Quasi-Objective Rating of Digital Samples

Last modified: 11.11.2019

This summer Henrik Strøm started his PhD at the Electronic Systems, Communication, Media, and Information Technologies (CMI). Henrik already knew CMI from taking educations in mICT and M.Sc.E. The PhD is arranged in collaboration with Copenhagen School of Design and Technology (KEA), where Henrik works as a lecturer teaching data science and software development.

The purpose of the PhD project is to research models that can do an advanced quality assessment of complex samples, and use these models to give immediate feedback to users, taking advantage of modern, powerful hardware. To do this, Henrik will create machine learning models (or other kinds of computational models) that align with a "public consensus" on the "quality" of samples, in the colloquial sense. Because this kind of quality assessment is intrinsically subjective, the first task is to collect ratings from real people, to establish a ground truth that can be used for training and validation of these models.

Initially, Henrik will work with simple 2D samples and later expand to more complex samples with a temporal factor. More specifically, he will start working with handwriting samples of the numbers 0 to 9, and then expand to Chinese characters that must be written with a particular order and direction of strokes. It is expected that some elements from these models can also be applied to data in various other domains.

Finally, to get a real-world evaluation of the models, Henrik will build the models into a tablet app that people, who are trying to learn writing Chinese characters, can use for training with a digital pencil. Such an app would allow for immediate feedback to the student and free up the teacher to focus on other learning topics.

During the PhD project, Henrik plans to find companies and universities in China for collaboration.