Effective Detection of Cassava Mosaic Disease (CMD) in Cassava Plants using DRNN Technique

Abstract

Convolutional neural networks (CNNs), in particular, have been utilized to make significant progress in the field of
automated leaf disease identification and diagnosis through the application of deep learning. The precision and efficiency of
agricultural processes have been considerably enhanced as a result of these improvements. The previous methods of disease
diagnosis relied on humanly generated features that were derived from photographs. This was the case throughout history.
Consequently, these strategies typically suffered from lower resilience and flexibility as a consequence of their actions. It
has been feasible to automatically extract features directly from raw images as a result of the transition to CNNs, which has
made it possible to overcome these limits. The goal of this research is to investigate the use of deep learning techniques for
the purpose of recognizing Cassava Mosaic Disease (CMD) in cassava plants as well as other ailments in tomato leaves. This
research is being conducted within the scope of this research. Certain deep architectures that integrate residual learning and
attention approaches were built in order to improve the process of extracting features from photographic images of leaves.
This was done within the context of improving the process. The experimental results on the Plant Village Dataset
demonstrated an impressive accuracy of 98% when it came to identifying diseases that affect tomato plants. These diseases
include early blight, late blight, and leaf mold. Experiments were conducted on the dataset. In a similar vein, a deep residual
convolutional neural network, also known as a DRNN, was able to achieve significant improvements in CMD detection as
compared to conventional CNN implementations. The utilization of block processing and various picture enhancement
techniques, such as decorrelation stretching and gamma correction, was the means by which this objective emerged. The
results of this study demonstrate the potential for CNNs to act as valuable tools for farmers, providing them with the
possibility to reduce crop losses and boost agricultural output. This potential was highlighted after the study was completed.
In order to further improve the practical usability of deep learning technologies in plant pathology, potential future research
areas may focus on scalability, model interpretability, and adaptability to a range of agricultural contexts. This is done with
the intention of further boosting the utilization of these technologies in plant pathology.