Real-Time Pest Detection Using ResNet-50 and Vision Transformer: An IoT-Enabled Mobile Application for Smart Agriculture

The advent of artificial intelligence (AI) has catalyzed a standard shift in smart agriculture, particularly in the domain of pest detection and disease classification. This study presents a novel hybrid deep learning framework integrating ResNet-50, Vision Transformer (ViT), and a custom convolutional neural network (CNN) within an IoT-enabled ecosystem to facilitate real-time, high-precision pest identification. The methodological pipeline encompasses data preprocessing, feature extraction through triple deep learning pathways, and fusion-based classification. The proposed model is rigorously validated against benchmark architectures, demonstrating superior accuracy and computational efficiency. The implementation within a mobile application underscores its practical applicability in precision agriculture, enabling farmers to undertake proactive interventions. The study delineates the limitations of existing methodologies and highlights the efficacy of multimodal AI techniques in transforming agricultural diagnostics. The results substantiate the robustness of the hybrid model in mitigating pest-induced agricultural losses, thereby fostering sustainable farming practices.