Object Recognition using Sparse Autoencoder with Convolutional Neural Network

摘要

Object recognition has turned into one of the demanding areas of exploration in the arena of image processing because of its uses in different applications such as Security, Robot navigation, Information retrieval, satellite imaging and various biometric applications. Several methods have been proposed which includes support vector machine, shape matching techniques, histogram techniques and various neural network techniques for feature classification and feature selection but have not been found to use autoencoder with CNN for object recognition. This paper enlightened the usage of unsupervised learning for pretraining purpose with the help of sparse autoencoder and use ConvNet with the aim to detect an object of each category, i.e., airplane, horse, automobile, bird, frog, cat, dog, ship, deer and truck etc. Autoencoder has two phase encoder and decoder. Encoder is used to encode the input image for extracting important features and decoder is used to restructure the input image. This paper shows the improved accuracy of CIFAR-10, CIFAR-100 and STL-10 dataset by using the proposed approach and also performing a number of cross-validation experiments on these object datasets.