Show simple item record

dc.contributor.authorDixit, Chahna
dc.date.accessioned2016-12-12T13:06:22Z
dc.date.available2016-12-12T13:06:22Z
dc.identifier.urihttp://hdl.handle.net/10222/72318
dc.description.abstractUnlike conventional handcrafted feature extractors, deep learning approach can extract generic image features without relying on explicit domain knowledge. Recently, there is a trend of combining handcrafted features with learned deep networks to leverage benefits of both. However, the usage of handcrafted features in existing methods are either by naïve concatenation or brute force from deep networks, and lack in addressing issues of parameter quality in the network. In this research, we propose a method that enriches the deep network features by injecting perceptual shape features - Generic Edge Tokens and Curve Partitioning Points, to adjust network’s internal parameter updating process. Thus, the modified convolutional neural network (CNN) produces image representation that tightly embraces benefits from both handcrafted and deep learned features. Our experiments on several benchmark datasets show improved performance compared to models using either handcrafted features or deep network representations alone, with reduced computation and faster convergence rate.en_US
dc.language.isoen_USen_US
dc.subjectDeep Learningen_US
dc.subjectConvolutional Neural Networksen_US
dc.subjectImage Classificationen_US
dc.subjectHandcrafted featuresen_US
dc.subjectPoolingen_US
dc.subjectNeural networks (Computer science)
dc.titleDiscriminative Shape Feature Pooling in Deep Convolutional Networks for Visual Classificationen_US
dc.typeThesisen_US
dc.date.defence2016-12-05
dc.contributor.departmentFaculty of Computer Scienceen_US
dc.contributor.degreeMaster of Computer Scienceen_US
dc.contributor.external-examinern/aen_US
dc.contributor.graduate-coordinatorDr. Malcolm Heywooden_US
dc.contributor.thesis-readerDr. Evangelos Miliosen_US
dc.contributor.thesis-readerDr. Dirk Arnolden_US
dc.contributor.thesis-supervisorDr. Qigang Gaoen_US
dc.contributor.ethics-approvalNot Applicableen_US
dc.contributor.manuscriptsNot Applicableen_US
dc.contributor.copyright-releaseNot Applicableen_US
 Find Full text

Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record