• Building a Pixel-Scan Camera From Scratch

  I'm partway through building a pixel-scan light field camera. So far I have a camera that captures perspective images by rotating a single photodiode with a pinhole lens across a scene. The images take an hour or more to capture at 64x64 pixels and they look like something out of an NES game.

• Reconstructing 1938 San Francisco from Aerial Images

  I worked with Justin Manley to use Harrison Ryker's 1938 set of aerial photographs of San Francisco to extract depth information and construct a point cloud of the city as it was in the '30s. We ran into some interesting challenges, from inaccurate GeoTIFF data to keypoint-snatching street labels. As the project currently stands, we have depth information for regions of the city covered by two or more photos, and plenty of room for improvement.

• Capturing Dynamic Baseline Stereo Video using Multirotor UAVs

  My thesis at Princeton University used multirotor drones to take stereo video from two different airborne perspectives. The captured video could be viewed through a VR headset to give the impression that the viewer was standing hundreds of feet tall, with eyes set far apart.

subscribe via RSS