Crop Disease Prediction Using Deep Learning Algorithms

2. Related Work

Many initiatives have been established to combat crop loss caused by diseases. Integrated pest management (IPM) methods have gained prominence over traditional insecticide usage in the past decade (Sumathi et al., 2022). Regardless of the approach, accurate disease identification at an early stage is crucial for effective plant disease management. Indigenous crop clinics, agricultural institutions, and associated agencies have historically played a role in disease identification. In recent times, the progress of such initiatives has been further augmented by the abundance of online resources and the easy accessibility of data for disease detection, capitalizing on the global surge in data consumption. Additionally, Android application-based tools have flourished, leveraging the widespread adoption of Android applications and their associated technologies (Sharma et al., 2020).

Smartphones, equipped with powerful high-resolution cameras and supported by efficient computing capabilities and fast processing speeds, present a distinctive platform for disease detection technology. Their inherent features, including high-resolution displays and built-in functionalities, enable innovative applications in the field of disease detection. The widespread adoption of smartphones worldwide has facilitated the potential reach of automated picture identification for disease diagnosis (S. Das et al., 2021). High-definition cameras, high-performance processors, and broad smartphone adoption are factors that contribute to the technical feasibility of such disease detection methods. In recent years, object identification and computer vision have made significant strides, with large-scale image recognition challenges based on the ImageNet dataset serving as a standard for various image recognition-related tasks, including object classification.

Advancements in deep convolutional neural networks have significantly reduced the error rate in image recognition tasks, reaching nearly 3.58 percent within the next three years (Saeed et al., 2021). Deep neural networks have been extensively employed for one-end-to-another-end learning across various domains. In the context of plant disease detection, neural networks establish a key-pair relation between input images, representing sick crops, and the corresponding outputs, indicating the specific crop and disease. Building a deep network requires careful consideration of its structure, operating parameters (nodes), and edge distortions (weights) to accurately map input values to desired outputs. The study of neural networks has seen substantial growth, driven by both logical and technical advancements (Nennuri et al., 2023).