Events & News
CS Colloquium
| Category | Lecture |
| Date | Tuesday, July 5, 2016 |
| Time | 11:00 am |
| Concludes | 12:15 pm |
| Location | Gould-Simpson 906 |
| Details | Please join us for coffee and light refreshments at 10:45am, Gould-Simpson, 9th Floor Atrium Faculty Host: Dr. Andrew Predoehl |
| Speaker | Luca Del Pero |
| Title | Ph.D. |
| Affiliation | Computer Vision Researcher at Blippar |
Discovering the physical parts of an articulated object class from multiple videos
I will present a motion-based method to discover the physical parts of an articulated object class (e.g. head/torso/leg of a horse) from multiple videos. The key is to find object regions that exhibit consistent motion relative to the rest of the object, across multiple videos. We can then learn a location model for the parts and segment them accurately in the individual videos using an energy function that also enforces temporal and spatial consistency in part motion. Unlike our approach, traditional methods for motion segmentation or non-rigid structure from motion operate on one video at a time. Hence they cannot discover a part unless it displays independent motion in that particular video. We evaluate our method on a new dataset of videos of tigers and horses, where we significantly outperform recent motion segmentation methods on the task of part discovery (obtaining roughly twice the accuracy). This work is part of our broader pipeline for learning properties of object classes from videos under weak supervision, of which I will also provide an overview.
Biography
Luca Del Pero is currently a computer vision researcher at Blippar. He has held a post-doctoral position in the CALVIN group at the University of Edinburgh. Luca obtained his Ph.D. at the University of Arizona under the advisement of Dr. Kobus Barnard.
His current research focuses on learning visual concepts from videos under minimal human supervision.
Luca has also worked on several other computer vision and machine learning problems, including 3D pose estimation, tracking, 3D scene understanding, automatic image annotation, Bayesian modelling and inference.