For a number of practical applications, it is desirable to have 3D models of the objects of interest. For example, if you wanted to have a custom shoe made, an app that takes a video of you foot and create a precise 3D model would make the process easier.
Another scenario is the modelling of non-rigid objects. For example, if we could induce from underwater videos of manta rays a 3D parameterized model of this species, the model would help in the re-identification of animals.
In this project, you will develop and implement with the help of the supervision team algorithms to induce 3D models from videos using techniques derived from 'structure from motion' and 'deep learning'. You will then run experiments to evaluate and compare the developed systems.
For a VRES project, only rigid objects will be considered and the aim would be to compare off-the-shelf systems based on 'Structure from motion'.
For research students, we will aim to develop solutions capable of modelling non-rigid objects. Deep learning techniques will be required.
Skills and experience
To be considered for this project, is is essential that you have excellent programming and mathematics skills. It is also desirable if you have prior knowledge of Python and machine learning.
You may be able to apply for a research scholarship in our annual scholarship round.
Contact the supervisor for more information.