A Critical Review of Object Detection using Convolution Neural Network

To recognize an object, for a human, is an easy task but for machines, to perform the same task with same efficiency is a complex task. For computer systems, images are sets of numeric values that have no meaning in itself. To make these numbers useful, diverse techniques have been proposed. Comparative to others, deep learning approaches achieved state-of-the-art performance in many computer vision applications, such as object detection, image classification, image retrieval, human pose estimation. To detect object of interest, Convolutional Neural Network (CNN) has been observed widely successful method. Few factors are there to get better accuracy and performance for instance efficient model, larger datasets and hardware support. This study aims to review CNN methods for object detection by highlighting the contribution and challenges from few recent research papers. Also how well to use these CNN techniques in combination to other methods for best suited results with other. Better performance such as increased accuracy, fast processing reduce error rates also introduced few new concerns and issues in parallel regarding the discussed methods such as time consumption, anonymous behavior of Neural Network. To address these issues a conceptual model is presented using CNN and Lease Square Support Vector Machine (LS-SVM).