Recommendation of Pesticide for Roof Top Pest Image Using Convolutional Neural Network Model

Urban Farm DB summary

Rooftop farming is gaining popularity for growing organic vegetables with minimal water use, but is highly vulnerable to pest infestation, and urban residents new to farming often fail to notice pest attacks. Noting that existing image-processing and machine-learning pest identification systems are typically disease-specific with low generalization accuracy and poor usability, this paper proposes a mobile-based pest identification system built on a pretrained convolutional neural network model (AlexNet). Various rooftop pests were evaluated experimentally using different kernel sizes and layer configurations, and the best-performing pretrained model was converted into a mobile application via a REST API to provide pesticide recommendations to novice users.