School of Computing Seminar with Joel Brogan, University of Notre Dame
"Tracing the Stories of Images on the Internet"
Everything on the internet these days borrows content from one source or another. Image cloning, splicing, and compositing techniques have become so advanced in such a short span of time, many classical digital forensic techniques for detecting modifications are no longer fully adequate to analyse forged imagery. Whether the purpose is for art, meme culture, or propaganda, it is important to be able to understand how a modified image came to be. The task of tracing the story of how images are uploaded, re-uploaded, filtered, and modified during their time on the web is called Image Provenance. If we can understand how sets of images are related to one another, or find the original images that a composite was based on, we can begin to gain insight as to why the image was modified. This talk highlights our group's efforts to build automatic tools that fetch composited images and trace their origins and modifications on a million to billion-image scale.
Joel Brogan is a PhD candidate from the Computer Vision Research Laboratory (CVRL) at the University of Notre Dame. His research lies in the area of machine learning and computer vision. He specializes in facial biometrics and image/video forensics, and has been funded by the DOD, DOE, and FBI for his past work. Currently, Joel is working with DARPA on new state-of-the-art methods for detecting and analyzing forged and tampered images and videos. He has published in many venues and journals, including the IEEE International Conference on Image Processing (ICIP) and Transactions on Image Processing (T-IP).
Friday, October 26 at 2:30pm
McAdams Hall, 119
821 McMillan Rd., Clemson, SC 29634, USA