Abstract
This study investigates the use of image processing techniques, particularly Convolutional Neural Networks (CNN), for identifying leafy diseases in Brassica Rapa (Chinese cabbage) cultivated in outdoor environments. Systematic experimentation determined optimal environmental parameters for image capture, including distance, angle, time, and brightness. Utilizing a diverse dataset of 3,296 categorized images, the researcher trained the CNN model to classify healthy leaves and those afflicted with back moth, leaf miner, and mildew diseases. The researcher used evaluation metrics such as accuracy, precision, recall, and F1 score to assess model performance, revealing an overall accuracy of 85.07 %, with notable success in identifying leaf miner-infected plants (85 % accuracy). This research underscores the importance of machine learning and image-based tools in sustainable agriculture, which enable early disease detection and reduce reliance on chemical control methods. The findings contribute to improved disease management practices and crop productivity, highlighting the integration technology into agricultural systems for more efficient farming.
