Rensselaer Polytechnic Institute & DualAlign
Plenary talk on Monday March 7, 4:15PM–5:15PM in Grandview Room
Abstract: This talk will describe the Image-Based Ecological Information System – the IBEIS project – a collaboration of teams from Princeton University, Rensselaer Polytechnic Institute, U. Illinois-Chicago, and the non-profit WildMe. Our goal is for IBEIS to be able to ingest many thousands of photos each day taken by field scientists, scouts, tourists, incidental photographers, camera traps and vehicle-mounted cameras, and process these images to detect animals, determine their species, and wherever possible identify the animals individually. A variety of computer vision algorithms, some novel and some adapted from existing work, have been applied to each step of this pipeline. Redundancy in the image collection and storage is important for improving algorithm effectiveness. Humans currently make all final decisions based on the results at each stage of the IBEIS computation, but more algorithm autonomy will be required as the system is scaled. The Wildbook information management system is being fully integrated into IBEIS to store and manage all ecological data, including non-image metadata. IBEIS has been used for large-scale citizen science events to count the plains zebras and Masai giraffes in Nairobi National Park and to provide a census of the endangered Grevy’s zebra throughout Kenya. It is in daily use at the Lewa Wildlife Conservancy. We are currently working to extend IBEIS to monitor the population and study the migration of humpback whales and of sea turtles. Plans are underway to apply IBEIS to large carnivores in east Africa and in southern and central Asia.
Biography: Chuck Stewart is Professor and Head of the Department of Computer Science at Rensselaer Polytechnic Institute in Troy, NY, and is also the founder of DualAlign LLC. He holds a BA in Mathematical Sciences from Williams College, and earned his MS and PhD in Computer Science at the University of Wisconsin, Madison. His research has been focused on computer vision for most of his career, including applications in industry inspection, medical imaging and remote sensing. He has served as associate editor of several journals, including IEEE Trans. on Pattern Analysis and Machine Intelligence, and has had a number of leadership roles in the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR). In 1999, along with two colleagues, he won the best paper award at CVPR. His research is now exclusively focused on developing and applying computer vision techniques to problems in ecology and population biology computation.
IARPA Program Manager
Plenary talk on Tuesday March 8, 4:15PM–5:15PM in Grandview Room
Title: Unconstrained Face
Abstract: The Intelligence Advanced Research Projects Activity (IARPA) invests in high-risk, high-payoff research to tackle some of the most difficult challenges of the agencies and disciplines in the Intelligence Community. The Janus Program funds research focused on unconstrained face recognition. Previously, face recognition research focused on identification in controlled settings such as mugshots, frontal viewpoints or where illumination, age and other physical attributes have little variation. Janus is seeking breakthroughs in computer vision, machine learning and other fields that may lead to significant advances in automatic detection and recognition of faces in natural settings. The goal goes beyond producing the next state-of-the-art in face recognition, but aspires to spark a deeper understanding of the underlying A.I. techniques behind face recognition.
This talk will give an overview of the IARPA Janus Program. It will touch on some of the techniques explored in the program including: geometric modeling, machine learning, and search and retrieval methods. Other areas of exploration include deep learning, hierarchical models and adversarial examples where face matching breaks down. A reference to other IARPA programs in computer vision will be made, as well as some broad categories for new research.
Biography: Dr. Terry Adams is a program manager for IARPA. Prior to joining IARPA, Terry worked for 16 years in the DoD. Terry worked for three years in operations, and then led a research effort to improve video processing. In 2011, Terry was named a Science & Technology Fellow by the Office of the Director of National Intelligence. His main areas of interest are computer vision, image recognition and large-scale processing of video. Terry maintains external research in the areas of statistical learning theory, probability and ergodic theory. He has publications in the Annals of Probability, the American Mathematical Society, IEEE Transactions and Ergodic Theory & Dynamical Systems.
Plenary talk on Wednesday March 9, 4:15PM–5:15PM in Grandview Room
Title: Fine Grained Visual Category Recognition and Perceptual Embedding
Abstract: In this talk I will provide an overview of my group’s research projects at Cornell Tech involving Computer Vision, Machine Learning and Human in the Loop Computing. Specific examples of projects we will cover include bird identification and learning perceptual embeddings of food.