Project Group: Seeing through disturbances

Topic: Seeing through disturbances

Assume you see an object under a water surface. The waves and disturbances on the water surface will distort your view of the object. Can we train a neural network to recognize or reconstruct the object nevertheless?

This project group combines aspects of Computer Vision and Computer Graphics. The group will start with a flat object, e.g. an image or a piece of text, at
the bottom of an aquarium. Then it will capture some real data with a camera, and also generate synthetic images with known ground truth using fluid simulation and rendering techniques (tools for this purpose are available). The group will use the resulting data to train deep neural networks for the purpose of getting an undistorted reconstruction of the object.

Participating in this project group requires team spirit, perseverance, and an interest in Visual Computing. You need to have passed Deep Learning and Computer Graphics I. Additionally, it is helpful to have participated in at least one other computer graphics course, e.g. Scientific Visualization or Computer Graphics III.

If you are interested, please contact Martin Lambers.