Classification of Diseases in Citrus Fruits using SqueezeNet

Recent advances in AI technology have paved the way to precisely detect and classify fruit and vegetable diseases. Vision based accurate detection and classification techniques can help farmers to quickly identify different types of plant and fruit diseases. Citrus fruits are globally famous due to their taste and nutrients. In this paper, a deep learning-based technique is proposed for classification of six different citrus diseases that severely affect the yield and quality of citrus fruits. Data augmentation is performed to enlarge the privately acquired dataset. Later the augmented dataset is used to retrain two lightweight pre-trained deep learning models; the SqueezeNet and MobileNetV2. Features are extracted from the retrained models which are subsequently optimized by using the Whale Optimization Algorithm. Classification results obtained from both the models show that the SqueezeNet model outperforms MobileNetV2 and attains an accuracy of 96%. The proposed citrus fruit disease classification technique also outperforms when compared with a current technique proposed in the literature