A Bayesian Approach to A Bayesian Approach to Unsupervised One- Unsupervised One -Shot Shot Learning of Object Learning of Object Categories Categories Li Li Fei Fei- -Fei Fei Rob Rob Fegus Fegus Pietro Pietro Perona Perona Presenter: Presenter: Tuyen Huynh Huynh – – Duy Duy Vu Vu Tuyen Slides from Li Fei-Fei http://www.vision.caltech.edu/feifeili/Fei-Fei_KUL2003.pdf 1
2
3
Summary Summary References References � Learning categories with one example is possible L. L. Fei Fei- -Fei Fei, R. Fergus, and P. , R. Fergus, and P. Perona Perona. A Bayesian approach to unsupervised . A Bayesian approach to unsupervised � � One One- -Shot learning of Object categories. Shot learning of Object categories. IEEE Inter. Conf. Computer Vision IEEE Inter. Conf. Computer Vision . . � Can reduce # of training example from ~300 to 2003 2003 L. Fei Fei- -Fei Fei, R. Fergus and P. , R. Fergus and P. Perona Perona. One . One- -Shot learning of object categories. Shot learning of object categories. L. � � 1~5 using prior from other categories IEEE Trans. Pattern Analysis and Machine Intelligence, Vol28(4), IEEE Trans. Pattern Analysis and Machine Intelligence, Vol28(4), 594 594 - - 611, 2006 611, 2006 Video of Fei Video of Fei- -Fei Fei’ ’s s talk at ICCV 2003: talk at ICCV 2003: � � http://www.robots.ox.ac.uk/~awf/iccv03videos/ICCV03 http://www.robots.ox.ac.uk/~awf/iccv03videos/ICCV03- -S11 S11- -P2 P2- -DIVX DIVX- -MP3.avi MP3.avi : http://www.variational http://www.variational- -bayes.org bayes.org/ / Variational Bayes Variational Bayes: � � 4
Recommend
More recommend