High performance vegetable classification from images based on AlexNet deep learning model

• DOI: 10.25165/j.ijabe.20181104.2690
• Publication Date: 2018-09-11T21:40:06Z
• Authors: Ling Zhu; ; Zhenbo Li; Chen Li; Jing Wu; Jun Yue
• Keywords: Overfitting; Dropout (neural networks); Data set; Contextual image classification; Sigmoid function

Deep learning techniques can automatically learn features from a large number of image data set. Automatic vegetable classification is the base many applications. This paper proposed high performance method for images based on deep framework. The AlexNet network model in Caffe was used to train set obtained ImageNet and divided into training test output function adopted Rectified Linear Units (ReLU) instead traditional sigmoid tanh function, which speed up network. dropout technology improve generalization model. extension reduce overfitting process. With different set, experimental results showed that accuracy decreases as decreases. verification indicated rate reached 92.1%, greatly improved compared with BP neural (78%) SVM classifier (80.5%) methods. Keywords: classification, learning, Caffe, Network, DOI: 10.25165/j.ijabe.20181104.2690 Citation: Zhu L, Li Z B, C, Wu J, Yue J. High Int J Agric & Biol Eng, 2018; 11(4): 217-223.