An Effective Framework for the Background Removal of Tomato Leaf Disease Using Residual Transformer Network

A precise and timely diagnosis of tomato plant diseases is essential for food security and sustainable agriculture. Using photographs of plant foliage and computer vision algorithms, the automated diagnosis of plant diseases has produced encouraging results. However, the accuracy of disease identification algorithms is frequently hindered by the complex and clogged backgrounds of these photographs. To address this issue, a transformer-based network-based classification model for tomato diseases is proposed. Real-time data is used to gather the initial, unprocessed images of tomato plants. Preprocessing was performed in order to eliminate the undesirable image pixels. In addition, the Fully Convolutional Network (FCN) is used to remove photograph backgrounds. The Residual Transformer Network (RTN) is ultimately utilized to classify maladies. Various metrics are employed to validate performance, which is contrasted with more conventional approaches. The RTN model classified tomato leaf diseases successfully. The accuracy analysis for the suggested tomato leaf classification model, RTN, demonstrated 8.04 percent better results than CNN, 6.81 percent better results than Res-net, and 4.44 percent better results than RNN with a higher background removal rate. The results demonstrate that an exceptional categorization rate is achieved to prevent a decline in agricultural output.