Visualization & Interactive Systems

We work on a diverse set of projects in visualization, graphics, and vision. Many of these projects are "applied" in the sense that we favor research which works toward real solutions to problems over research which is only of interest to academia. Examples of current projects include:

• Rip currents - We are working on computer vision methods for rip current detection, towards practical visualizations for letting everyday people avoid dangerous ocean conditions. 

• ML assisted data collection - We are working on tools to make collecting scientific datasets faster. For example, flying drones with computer vision assistance to detect when and where to gather data along a planned path. 

• Engaging citizen scientists across disciplines - We are working on tools that allow scientists from different fields to quickly create applications they can use for data collection by citizen scientists without having to write a line of code.  The apps can also be equipped with ML assist feature.

• Improved human annotation - Labeled data sets are at the core of much of modern ML. We are working to improve choices on which data is labeled, how we know when we have enough, how to deal with limited or noisy training data, and how to improve correctness of labels.

• Synthesize data augmentation - Training data are often collected under good conditions.  However, real world application of ML models must also perform in less than ideal conditions.  Collecting and labelling training data in such conditions may not be possible, costly, and time consuming.  We are working on synthesizing training data that simulates different weather (e.g. fog, rain, snow) and environmental (e.g. glare, shadows, occlusions) conditions.

• 3D face reconstruction - The human face is perhaps the most photographed object on the planet. However modern mobile phones still do not provide a 1mm accurate 3D model to go along with all those selfies. Enabling this will allow better photo editing, avatars for 3D worlds, and measurements for fitting your favorite custom eyeglasses and VR headsets.

• Vehicle safety - Modern cars have systems for blind spot warning, parking, and driver monitoring. We are working on the computer vision which powers all these systems.

• Navigation aid for visually impaired people - using augmented reality headset and computer vision techniques to reconstruct 3D scenes.  The scene, pose information from headset and/or camera is mapped to spatial audio to help visually impaired people avoid static and dynamic obstacles.