Unified Multi-Scale Feature Network (UMFN) for Accurate Detection and Classification of Cauliflower Diseases

Abstract

The detection and classification of cauliflower diseases are crucial for crop health as well as the maximization of agricultural yield. In this paper, a new deep learning architecture, the Unified Multi-Scale Feature Network (UMFN), is proposed that has the potential to achieve state-of-the-art disease classification and detection accuracy on cauliflower crops. The novel model integrates a unique integration of multi-level attention mechanisms, intensity-feature locators, and advanced multi-scale feature extraction technologies for enhanced detection precision. By integrating a resized Integrated Feature Pyramid Network (IFPN) architecture, the model enhances lesion localization precision in different scales while maintaining critical spatial information. Large-scale experiments confirm that UMFN surpasses current state-of-the-art models on primary performance measures such as precision, recall, and mean Average Precision (mAP). The results emphasize the potential of the model to enhance precision agriculture through the provision of a robust, effective, and automated system for real-time disease monitoring of cauliflower crops. This paper focuses on the revolutionary significance of deep learning in agriculture through the provision of a way for enhanced disease control, maximum crop yields, and global food security.