NTEGRATED DEEP LEARNING FRAMEWORK FOR DISEASE CLASSIFICATION IN AGRICULTURAL COMPUTER VISION

This study investigates the use of computer vision in agriculture, specifically focusing on classifying maize diseases. It explores the integration of machine learning and deep learning techniques, recognizing the limitations of using machine learning alone. The research highlights the need for additional image processing steps like preprocessing, feature extraction, segmentation, and augmentation to improve classification accuracy. The approach centers on the ResNet50 model, a deep learning architecture designed for automatic feature extraction. While deep learning excels at handling large datasets, it requires significant computational power. The proposed methodology employs a holistic preprocessing pipeline, beginning with image segmentation to convert raw images into labeled regions for detailed analysis. Data augmentation, using the U Square Net model, is applied before splitting the dataset to ensure an unbiased model evaluation. Techniques like Gaussian blur and normalization further enhance input features for the neural network. The model, built using PyTorch, combines the pre-trained ResNet50 model for feature extraction and effective pattern recognition. The study concludes with a thorough evaluation of the model’s performance, demonstrating its potential to improve maize disease classification. This research offers a comprehensive framework for accurate disease identification in agricultural settings through the synergy of image processing and deep learning.