Evolving Deep Neural Networks for Efficient Image Classification

Abstract

Image classification or recognition has become base for many real time applications like: object recognition, object tracking, action recognition, drug recovery, video tracking/recognition etc. These applications accuracy depends on the results of image classification system. In literature many researchers have used low level features such as Texture, Gabor, Gradient, SURF and high level features such as CNN for image classification. In recent years, Deep Neural Networks are widely used for image analysis due to its deep feature estimation procedure (i.e. the more no. of features leads to good classification). Thus, this paper proposes three different models based on deep neural networks for efficient Image Classification. The proposed methods are focused on reducing the error rate while increasing the classification accuracy (above 95%).