This page is out of date and will be updated sometime in the future.
My research interests are related to image processing, computer vision, and machine learning. I’m especially interested in problems related to semantic image segmentation and scene labeling, weakly-supervised learning, and applications related to subsurface imaging.
Computational Seismic Interpretation
Our work in computational seismic interpretation is mainly concerned with developing computational methods and algorithms to help seismic interpreters and geophysicists process and understand seismic data more efficiently. I've mainly worked on developing new seismic attributes, detection and delineation of subsurface structures, developing expressive descriptors for seismic data, and using these descriptors to detect and classify certain structures in seismic volumes.
An important step towards image understanding (with its widely useful applicaitons) is scene labeling (sometimes referred to as "semantic segmentation", or "scene parsing"). Scene labeling is the task of assigning every pixel in an image with a label describing the category to which the pixel belongs. Scene labeling has many exciting applications, such as enabling autonomous vehicles to navigate the streets and watch out for pedestrians, or improving automated image annotation systems.