As someone who has been working in computer vision and pattern recognition for over three decades, I have watched with interest how most existing efforts in computer vision are based on pattern recognition methodologies. More and more, the algorithms take the form, data (image, video, depth, etc.), features (SIFT, Hog, LBB, attributes, dictionaries, etc.) followed by a favorite version of SVMs. This approach has generated successful algorithms such as deformable parts model for object detection, attribute-based face verification, etc. More recently, a different manifestation of pattern recognition algorithms, based on deep learning has produced best results on ImageNet and LFW data sets. While I am a devoted student of pattern recognition school from Purdue, I would like to argue that domain shifts due to illumination and pose variations, blur and resolution as well as occlusion will require the incorporation of models and geometry to realize generalizations across data and help design robust systems. I call for a balanced approach that effectively combines imaging and geometric models and data for reaping long term gains. [Go to the full record in the library's catalogue]
This video is presented here with the permission of the speakers.
Any downloading, storage, reproduction, and redistribution are strictly prohibited
without the prior permission of the respective speakers.
Go to Full Disclaimer.
This video is archived and disseminated for educational purposes only. It is presented here with the permission of the speakers, who have mandated the means of dissemination.
Statements of fact and opinions expressed are those of the inditextual participants. The HKBU and its Library assume no responsibility for the accuracy, validity, or completeness of the information presented.
Any downloading, storage, reproduction, and redistribution, in part or in whole, are strictly prohibited without the prior permission of the respective speakers. Please strictly observe the copyright law.