Plant Disease Identification Using a Dual Self-Attention Modified Residual-Inception Network

The early detection of plant diseases reduces agricultural loss. In the field of computer vision and pattern recognition, deep learning (DL) techniques, particularly convolutional neural networks (CNNs), are widely employed. To identify plant diseases, researchers put forth various DL models. However, DL models require many parameters to learn the underlying patterns of the plant disease, increasing training time and making it challenging to deploy on small devices. This study introduces a novel DL model utilizing a dual self-attention modified residual-inception network (DARINet), which integrates the multi-scale, self-attention, and channel attention features with the residual connection. The proposed approach is evaluated on two plant disease datasets such as Cassava and Rice leaf, achieving an accuracy of 77.12% and 98.92%. In Comparision to state-of-the-art DL models, our proposed approach attains higher accuracy with fewer parameters.