Image Recognition with Deep Learning Techniques

This paper investigates a Deep Learning (DL) approach for image recognition. We have considered two DL neural models: Convolutional Neural Network (CNN) and Deep Belief Network (DBN). We have chosen several architectures for each of the proposed models. We have chosen Caltech101 dataset to train and test the above proposed models; this database is composed by images belonging to 101 widely various object categories. One has considered the SVM-KNN algorithm as benchmark, being the best available model selected by the Caltech101 database issuer. Several dataset preprocessing techniques are considered. Using our proposed DL approach, we have obtained a correct recognition score of 67.23%, corresponding to an increase of 1% over the chosen benchmark algorithm (66.23%). Key-Words: Deep Learning (DL), Convolutional Neural Networks (CNN), Deep Belief Networks (DBN), image classification, Caltech101