Abstract: The primary causes of the significant decline in grape yield are grape diseases. Therefore, the development of an automatic grape leaf disease identification system is imperative. The remarkable results that deep learning techniques have lately obtained in a variety of computer vision challenges motivate us to apply them to the issue of identifying grape illnesses. This paper proposes an integrated method-based architecture for convolutional neural networks (CNNs). The suggested CNN architecture, or UnitedModel, is made to differentiate between healthy leaves and leaves that have common grape diseases including black rot, esca, and isariopsis leaf spot. The suggested UnitedModel can extract complementary discriminative features because it combines multiple CNNs. As a result, UnitedModel now has better representation. Using the withheld PlantVillage dataset, the UnitedModel has been assessed and contrasted with multiple cutting-edge CNN models. Based on multiple evaluation metrics, UnitedModel performs the best, according to the experimental results.
Keywords: Grape leaf disease, image processing, feature extraction.
|
DOI:
10.17148/IJARCCE.2024.134176
