AN OPTIMIZED APPROACH FOR INTRA CLASS FRUIT CLASSIFICATION USING DEEP CONVOLUTIONAL NEURAL NETWORK
Rishipal Singh, Rajneesh Rani and Aman Kamboj
Fruits classification is one of the influential applications of computer vision. Traditional classification models are trained by considering various features such as color, shape, texture, etc. These features are common for different varieties of the same fruit. Therefore, a new set of features is required to classify the fruits belonging to the same class. In this paper, we have proposed an optimized method to classify intra-class fruits using deep convolutional layers. The proposed architecture is capable of solving the challenges of a commercial tray-based system in the supermarket. \textcolor{red}{As the research in intra-class classification is still in its infancy, there are challenges that have not been tackled. So, the proposed method is specifically designed to overcome the challenges related to intra-class fruits classification. The proposed method showcases an impressive performance for intra-class classification, which is achieved using a few parameters than existing methods. The proposed model consists of Inception block, Residual connections and various other layers in very precise order.} To validate its performance, the proposed method is compared with state-of-the-art models and performs best in terms of accuracy, loss, parameters, and depth.
© Rishipal Singh