A unique ADAGRAD optimized DCNN with RESNET-18 Architecture for Indoor Agriculture-Based Crop Yield

A new kind of cityscape has evolved in the so-called “megacities” in the last few decades. By 2030, experts predict that over five billion people will call cities home. The World Health Organization estimates that by 2050, there will be 10 billion people on Earth, with 6.5 billion of them residing in metropolitan areas. To fulfill this demand, food production must grow by 70%. In order to facilitate the transition to more sustainable and efficient food supply solutions and the transportation of clean, fresh vegetables, several initiatives have been launched in megacities to bolster the new ecosystem services. An unusual agricultural technique that may address food insecurity, indoor agriculture (IA) consumes less area, provides better yields, and is resource-efficient. Crop monitoring, biotic and abiotic stress detection, irrigation, microclimate prediction, energy efficient controls, and crop growth prediction are just a few of the many uses of deep learning (DL) that have recently been brought to IA. But there hasn’t been a comprehensive evaluation of DL’s capabilities to address a wide range of IA issues. The setting, such as the circumstances within a greenhouse, is crucial in agriculture. A number of recent studies have shown that farmers might benefit from data, reminders, and warnings provided by sensors that track productivity or by cameras that capture information on agriculture. Using an improved RESNET-18-based DCNN model with an Adagrad optimizer, we provide a new approach to indoor tomato yield prediction in this research. The suggested model outperforms the state-of-the-art models and demonstrates a considerable improvement in prediction accuracy.