Plant Health and Severity Recognition System for Pechay (Brassica rapa ssp. chinensis) Using Convolutional Neural Network

Pechay, known as Bok Choy, is essential in Filipino cuisine, producing over 47,500 metric tons in 2020. Health monitoring is recommended to enhance pechay yield and quality for easier application of mitigation methods upon health declines. This study developed a system to help local farmers or hydroponics enthusiasts identify pechay leaf symptoms, allowing timely interventions. Health monitoring involves visual inspections for symptoms like leaf yellowing (chlorosis), browning, or blacking (necrosis). A REST API-based client (Raspberry Pi) and server (laptop) system automates this process. The model outputs the (1) health condition and (2) severity of the pechay. It employs a multioutput convolutional neural network (CNN) model trained with a self-collected dataset from a hydroponic setup. The model reached 89% and 88% accuracy for plant health and severity detection, respectively. The system, tested on over 100 pechay samples, achieved an 82% accuracy rate. Chi-squared testing showed no significant difference between the system’s assessments and actual plant health conditions.