Classification of Agricultural Leaf Images using Hybrid Combination of Activation Functions

Agriculture is the primary source of India for the raise of livestock and it's the backbone of the economy that gives surplus food materials for living and it conjointly acts as stuff for industries. India's twenty eighth percent of the national income depends on the agricultural sector. Agriculture contributes to the expansion rate of employment. The most important issue faced in agriculture is loss or damage of crops due to environmental conditions such as flood, lack of water during summer and various pests which affects the crops. Among these crop diseases and pests have an enormous contribution to agricultural huge loss. The diseases can be identified with advanced deep learning algorithms. Though many DL methods works well in identifying the disease still there are misclassifications during training. To avoid the misclassification in predicting the appropriate disease an ensemble activation approach is proposed. In this paper, deep learning based technique is enforced for the automatic detection of plant leaf diseases. The convolution Neural Network approach is applied during this paper. Activation functions play a vital role in convolution Neural Network (CNN). In this paper, a combination of hybrid activation function is proposed to improve the accuracy rate of the CNN model. The hybrid Activation function is tested and trained on different datasets. The hybrid activation function is compared with the standard activation function ReLU. The Experimental results by trial and error show that the hybrid activation function achieves higher accuracy than the ReLU activation function.