Vision and Autonomous Systems Seminar
- Newell-Simon Hall
- JAYAKORN VONGKULBHISAL
- Ph.D. Student
- Department of Electrical and Computer Engineering
- Carnegie Mellon University / Instituto Superior Técnico, Portual
Discriminative Optimization (DO): Theory and Applications to Computer Vision
Many computer vision problems are formulated as the optimization of a cost function. This approach faces two main challenges: (i) designing a cost function with a local optimum at an acceptable solution, and (ii) developing an efficient numerical method to search for one (or multiple) of these local optima. While designing such functions is feasible in the noiseless case, the stability and location of local optima are mostly unknown under noise, occlusion, or missing data. In practice, this can result in undesirable local optima or not having a local optimum in the expected place. On the other hand, numerical optimization algorithms in high-dimensional spaces are typically local and often rely on expensive first or second order information to guide the search. In this talk, I will present our work on Discriminative Optimization (DO), a method that learns search directions from data without the need of a cost function. In essence, DO explicitly learns a sequence of updates in the search space that leads to stationary points that correspond to the desired solutions. While the concept of DO resembles that of the supervised descend method (SDM), we provide an analysis on the convergence of the training error and its relation to convexity, which leads to a framework for deriving feature functions for different problems. We demonstrate the potential of DO in the problems of 3D point cloud registration, camera pose estimation, and image denoising, and show that DO can match or outperform state-of-the-art algorithms in terms of accuracy, robustness to perturbations, and computational efficiency.
Jayakorn Vongkulbhisal received his BEng degree from Chulalongkorn University, Thailand, and MSc degree in ECE from Carnegie Mellon University in 2011 and 2016, respectively. He is currently working toward his PhD degree at CMU and IST-Lisbon under the supervision of Prof. Fernando De la Torre and Prof. Joao Paulo Costeira. His research focus is in optimization and machine learning for computer vision.
Sponsored in part by Disney Research