10-701 Student Presentation - Scalable Support Vector Machines and applications to Computer Vision

Wen-Sheng Chu, Ishan Misra, Francisco Vicente

Support Vector Machines (SVMs) are powerful tools in both classification and regression tasks. The computer vision community, in particular, relies heavily on SVMs for object detection [Felzenszwalb et al. 2010], image retrieval [Shrivastava et al. 2011], face image analysis [Kumar et al. 2009], etc. In thisproject we address the problem commonly faced while using SVMs on large datasets. Specifically, we focus on the scalabilities of Exemplar-SVMs and nonlinear SVMs with RBF, X^2 and intersection kernels. The effectiveness of our approaches are demonstrated on two popular computer vision problems, i.e., object detection and facial attribute recognition.