Research Symposium >
SLIIT Student Research -2016 >

Please use this identifier to cite or link to this item:

Title: Deep Object Vision Convolutional Neural Network
Authors: Karunarathne, P.G.U.D.M
Bamunusinghe, J.
Keywords: Convolutional Neural Network (CNN
Object Recognition
Data Whitening,
Issue Date: 2016
Publisher: SLIIT
Abstract: Deep learning has achieved great heights in terms of accuracy in large-scale image classification task in the recent past with the introduction of Convolutional Neural Network (CNN) model. We propose an improved version of CNN model based on the Alexnet, called Deep Object Vision (DeepOV). The proposed model was designed and implemented in Python environment using Theano backend and Keras libraries for the efficient GPU utilization. DeepOV model uses the CIFAR-10 dataset for training testing and validation purposes and demonstrates better accuracy rates in comparison to the classical object recognition models. We achieved 80% validation accuracy on test set, overwhelming the common drawbacks that exist in Alexnet model. The proposed DeepOV model has 26 million parameters including four convolutional layers, some of which are followed by max-pooling layers, and two fully connected layers and finally a softmax layer. We introduce data preprocessing, image whitening and regularization techniques to further improve the recognition accuracy of the proposed model.
Appears in Collections:SLIIT Student Research -2016

Files in This Item:

File Description SizeFormat
Deep Object Vision Convolutional Neural Network (2).pdf609.61 kBAdobe PDFView/Open
View Statistics

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.


Valid XHTML 1.0! DSpace Software Copyright © 2002-2010  Duraspace - Feedback