Crop disease detection using EfficientNetB0 deep learning approach for precision agriculture

Abstract

Crop disease detection plays a crucial role in modern agriculture by enabling timely interventions to address diseases that directly impact crop yields and global food security. This research aims to explore a novel deep-learning approach for detecting crop diseases. We utilize the EfficientNetB0 architecture, specifically designed to extract intricate features from large datasets, making it well-suited for identifying various crop diseases. The model is fine-tuned through transfer learning to adapt to the specific characteristics of crop diseases, enhancing its classification performance. Using a dataset of approximately 87,000 images across 38 classes, our findings demonstrate that this model is highly effective in detecting subtle patterns and features associated with crop diseases. It achieved 99.51% test accuracy with a loss of 0.0165, underscoring its practical efficacy. Such a model could be applied in precision agriculture field monitoring systems, assisting farmers in managing their crops more effectively. Early detection of crop diseases provides an advantage to farmers, allowing them to respond quickly to minimize losses. This not only supports early detection but also contributes directly to sustainable disease management and yield protection. Models like EfficientNetB0 empower farmers to grow crops more efficiently and maximize their yield potential, as the world will need to increase food production by 70% by 2050 to accommodate a growing population. This research underscores the importance of utilizing advanced algorithms based on deep learning to identify crop diseases and highlights the transformative impact of technology on modern agriculture. Integrating the EfficientNetB0 model into agricultural workflows represents a significant step toward building resilient and sustainable food systems.

Abstract from the paper’s OpenAlex record; open copy licensed cc-by-nc-nd. Read at the source ↗

The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.