Plant Disease Classification using Interpretable Vision Transformer Network

Agriculture is the backbone of the Indian economy and it caters to the basic necessity of food for billions. Hence, increasing the production yield is a serious challenge, however, it sometimes gets affected with the microorganism-caused infections that severely affects the per acre produce. Therefore, the objective of this study is to develop an automated system for an early detection of plant disease. In the proposed work, pre-trained Vision Transformer architecture has been fine-tuned for plant disease classification. The classification decision made by the proposed model has also been interpreted using the GradCAM algorithm with the help of visualisation. The performance of the proposed method has also been compared with the state-of-the-art pre-trained convolution neural networks fine-tuned for the same purpose. The proposed method has been tested with the 'PlantVillage' public dataset which consisting of 39 classes of plant images. The experimental results show that the proposed method classifies the 39 classes (38 diseased/healthy, 1 leaf image without background) of plant images with 98.22% accuracy.