An Efficient Tomato Leaf Disease Classification Framework by Background Removal Using Fully Convolutional Network and Residual Transformer Network

Tomato is a widely utilized vegetable that secured a superior place in enhancing the economic rate in agriculture. Production quantity of tomatoes secured seventh place worldwide. Still, leaf diseases attained in the tomato plant are considered a more challenging issue as they affect the normal growth of the plant badly. Here, different kinds of plant diseases have occurred in the plant leaves which generate more losses in crop yield. In the early phase, the accurate prediction of plant disease effectively minimizes the production loss and offered to enhance crop yield. Moreover, the conventional approaches faced more complexity as the existing system needs more computational time and they are costly. To overcome this problem, an effective tomato disease classification model is proposed using a transformer-based network. Initially, the raw tomato plant images are gathered through real-time data. Then, they have undergone a pre-processing stage to remove the unwanted image pixels. Further, the backgrounds of images are removed by using the Fully Convolutional Network (FCN). Finally, the disease classifications are accomplished by using Residual Transformer Network (RTN). The performance is validated with divergent measures and compared with traditional methods. Thus, the results declare that it achieves an impressive classification rate to avoid less crop productivity.