Advantages And Disadvantages Of Cnn Models

1361 WordsAug 15, 20176 Pages
In everyday life, humans easily classify images that they recognize e.g. surrounding objects are easily identified, but the classification of images in the disease detection, video surveillance,vehicle navigation is a challenging and important task. To mitigate the risk,computers are trained to classify images using the various algorithm ~citep{kamavisdar2013}. Image Classification using CNN model is widely used as they are powerful in achieving high accuracy with minimum error rate. CNN’s have shown remarkable classification results using standard architectures ~citep{krizhevsky2012,simonyan2014,zeiler2014,szegedy2015} this is complex to understand and implement, hence ~citep{hasanpour2016,wang2016} expresses the need to develop CNN…show more content…
followed by Literature Review that discusses the design of various architectures and study of parameter values, the steps used for the research is discussed in the methodology section, followed by implementation section where the techniques used for implementing the model is highlighted, Evaluation section includes different case studies to evaluate the performance of the model.The paper ends with Conclusion, abbreviation and references used for study in earlier sections. subsection{Research Objective} This research is focused on finding the combination of parameters ( use of different activation functions, drop-out values, filter size, batch size,layers) that will improve classification accuracy of the network. subsection{Hypothesis} Using a good combination of parameter values will help to improve the classification and overall accuracy of the network. subsection{Organization Of study} subsubsection{Motivation} egin{enumerate} item Classification using Machine learning is one of the most important fields of research and development these days with CNN proven to be effective not only in Image Classification but also video, pattern and face recognition. ~citep{wang2016,egmont2002} item It is one of the most challenging fields, as understanding the Convolutions on the images will help to increase knowledge in this domain which can then be later used with other applications. item Due to large number of parameters it becomes difficult for the
Open Document