|
Jon Barron I am a fifth year PhD candidate in EECS at UC Berkeley. My advisor is Jitendra Malik. I am funded by the NSF GRFP. |
 |
I'm interested in computer vision, machine-learning, and computational photography. Most of my research is about figuring out the physical world (shape, paint, light, etc) that created a single image. I also work in astronomy and biological image analysis.
 |
By embedding mixtures of shapes & lights into a soft segmentation of an image, and by leveraging the output of the Kinect, we can extend SIRFS to scenes. |
 |
Boundary cues (like occlusions and folds) can be used for shape reconstruction, which improves object recognition for humans and computers. |
 |
We present SIRFS (shape, illumination, and reflectance from shading), the first unified model for recovering shape, chromatic illumination, and reflectance from a single image. |
 |
Given just a single grayscale image of some masked object, we estimate the shape, albedo, and illumination that created that image. We outperform all previous grayscale "intrinsic images" algorithms. |
 |
We present a large RGB-D dataset of indoor scenes and investigate ways to improve object detection using depth information. |
 |
To address shape-from-shading and intrinsic images simultaneously, we impose "naturalness" priors over albedo and shape and recover the most likely explanation of a single image. |
 |
We introduce a model and feature representation for joint texture classification and segmentation that learns how to classify accurately and when to classify efficiently. This allows for sub-linear coarse-to-fine classification. |
 |
Using known catalog proper motions, we can accurately estimate the date of origin of historical imagery given only raw pixel data. |
 |
We use computer vision techniques to identify and remove diffraction spikes and reflection halos in the USNO-B Catalog. In use at Astrometry.net |
 |
|
 |
(Erdös = 3)