Mush-Room for Innovation: Internet of Things, Machine Learning, and Generative Pre-Trained Transformers Integration

Mushroom farming is a vital sector in Jordan's agriculture (set to reach 1,060 metric tons by 2026). This study aims to revolutionize mushroom farming through the integration of the Internet of Things (IoT) and artificial intelligence (AI) capabilities through machine learning and the generative pre-trained transformer-4omni (GPT-4o) most recent model. A comprehensive hardware and software architecture was developed after training three machine vision models (Visual Geometry Group 16 - VGG16, Inception Third Version - InceptionV3, and Residual Network - ResNet50) and integrating GPT-4o for sound and text recommendations and an interactive dashboard with the alarming system. A robust dataset of (n = 4,755) images was meticulously processed, achieving the highest classification accuracy of 91.5% with ResNet50. The system’s static and real-time vision capabilities were validated with a prototype, integrating sensors for temperature, humidity, and gas levels. Unlike previous studies, this research uniquely combines IoT and AI with GPT-4o, providing intelligent recommendations and visual dashboards. The results demonstrated significant improvements, with practical applications extending to open-source tools for global use. Future research could explore drone integration and address privacy and security concerns. This study sets a new standard in agricultural technology, offering a scalable, efficient solution for modern mushroom farming and the potential to extend to further applications.