Fruit Classification Based on Six Layer Convolutional Neural Network

Siyuan Lu, Zhihai Lu, Soriya Aok, Logan Graham

Abstract: Automatic fruit classification is a difficult problem because there are so many types of fruits and the large inter-class similarity. In this study, we proposed to use convolutional neural network (CNN) for fruit classification. We designed a six-layer CNN consisting of convolution layers, pooling layers and fully connected layers. The experiment results suggested that our method achieved promising performance with accuracy of 91.44%, better than three state-of-the-art approaches: voting-based support vector machine, wavelet entropy, and genetic algorithm.

External IDs:dblp:conf/icdsp/LuLAG18