MULTI LABEL FOLIAR DISEASE CATEGORIZATION IN APPLE ORCHARDS USING DEEP LEARNING

Abstract

The accurate detection and classification of foliar diseases in apple orchards are crucial for ensuring crop health and optimizing yield. Traditional methods of disease identification, which rely heavily on visual inspection by experts, are not only time- consuming but also prone to errors and lack scalability. This research introduces a robust deep learning model leveraging the MobileNetV2 architecture, enhanced with custom layers and advanced data processing techniques, to address these challenges through automated multi-label classification of foliar diseases. Utilizing the FGCV 2021 dataset, comprising over 23,000 high-resolution images labeled for multiple diseases, the model demonstrates superior performance in recognizing complex disease patterns under varied environmental conditions. The study employs extensive data augmentation techniques including rotations, flips, and color adjustments to enhance model generalization across different lighting and background scenarios. The training process is optimized through adaptive learning rate adjustments and early stopping mechanisms to prevent overfitting, thus ensuring the model's robustness. Performance evaluation metrics such as accuracy, precision, recall, and F1-score consistently indicate that the proposed model outperforms existing benchmarks, achieving an accuracy of 92.46%, and an F1-score of 90.87%. The implications of this research are significant, offering a scalable and efficient tool for real-time disease monitoring in apple orchards. This not only aids in timely and effective disease management but also potentially reduces economic losses and improves agricultural sustainability. Future work will focus on expanding the dataset to include more diverse environmental conditions and exploring real-time deployment scenarios to enhance the practical applicability of the model in precision agriculture. Another area could be adapting this solution to a wider range of crops and diseases, as well as integrating with other evolving technologies like IoT or drones for effective and larger-scale crop monitoring.