Our lab builds publicly available tools for geolocating images and validating their metadata (i.e., GPS or timestamp). We produced and maintain geocalibration.org, a 2d camera calibration tool, and projectlive3d.com, a 3d version, and are in the process of building the web version of a MATLAB tool I wrote for validating image metadata from the shadows in a scene.
rePhoto, available for iOS and Android devices, is an image capture application explicitly designed to support repeat photography -- the process of taking a new image from exactly the same perspective as a previous image. In rePhoto this is made easier by showing the previous picture half see-through so that a new picture can be more accurately aligned. The app supports both personal repeat photography as well as large scale, citizen science projects archived and visualized at Project rePhoto.
We study natural phenomena like shadows, clouds, and reflections and characterize how they change at different places, different times and different camera settings. Understanding these constraints is important when trying to answer forensic questions about images such as exactly when or where an image was taken.
The Archive of Many Outdoor Scenes (AMOS) dataset contains 579,881,580 images taken from 23343 webcams (and counting) located around the world. This dataset is unique in that it contains images from significantly more scenes than in previous work. The cameras in the dataset were selected by a group of graduate and undergraduate students using a standard web search engine. Images from each camera are captured several times per hour using a custom web crawler that ignores duplicate images and records the capture time. One impetus for collecting these images was to provide a dataset to enable empirical assessment of our ideas regarding the statistics of natural scenes captured from static cameras.