Disease Detection in Tomato Plant using Deep Residual Learning

With the accelerated rise of human population, food is bound to be quintessential of our worries in the future, thus mitigating plant diseases, and errors in their detection would be extremely helpful. The traditional method of examining plant diseases involved an erudite of the field to be present, but various developments in neural networks have made this process trivial by requiring only the image of the crop for its disease detection. There has been a myriad of attempts to improve the scope of plant disease detection to try and make it a lot more computer-dependent and mitigate the involvement of the end-user by using deep learning. This is achieved by using convolutional neural networks which excel in the field of image detection. This paper is based on the fine-tuning aspect of deep learning, i.e., this paper tries to improve the performance of a pre-trained model and compare the accuracies of preexisting models which are popular in this aspect of plant disease detection. This model has been trained specifically for tomato plant leaf dataset, which was sourced from Kaggle, which includes 10 distinct classes including both diseased and healthy leaves.