Detection of calcium deficiency and physiological status in strawberry leaves using deep learning

Detecting diseases, disorders and physiological changes in plants is important to prevent crop damage and increase food production. In this work, a CNN network is proposed to detect calcium deficiency and a physiological change related to chlorophyll deficiency in strawberry plant. Transfer learning is applied to some common benchmark models to compare their performance with the proposed model. A dataset of images of strawberry leaves was collected from a greenhouse and used to train the models. Moreover, the dataset contains 1955 images for 3 different classes that are "healthy leaves", "leaves with calcium deficiency" and "old leaves" (lack of chlorophyll). The classification was done in two stages, a binary classification in which healthy leaves and leaves with calcium deficiency were used, the second stage involved multi-class classification after adding old leaves. It has been shown that the proposed model achieved an accuracy of 98.97% which is higher than transfer learning models which are used. In addition, effect of learning rate on the performance of the proposed model was discussed. Keywords: Deep leaning, Convolutional neural network, Strawberry disorder, Artificial intelligence, Machine learning