IoT-Based Plant Growth Chamber with YOLOv8 for Anthracnose Disease Severity Classification in Chili Pepper

Abstract

Plant growth chambers provide controlled environments for agricultural research, enabling precise monitoring of crop diseases under optimal microclimate conditions. This paper presents an integrated IoT-based smart plant growth chamber system utilizing YOLOv8 machine learning for the automated classification of anthracnose disease severity in chili peppers (Capsicum annuum L.). The system integrates multiple subsystems, including environmental control, a robotic camera with 2-axis movement, a gateway for data communication, and remote monitoring capabilities through a cloud server and a web interface. Dataset labeling was performed using LabelImg and Roboflow, with data augmentation increasing training samples from 70% to 86%. Three YOLOv8 models were evaluated: YOLOv8L (150 epochs), YOLOv8N (100 epochs), and YOLOv8N (398 epochs). Based on our test so far, the YOLOv8L model achieved the best performance with mAP of 67.4% and successfully detected 44 out of 102 test samples (43% detection rate) across multiple disease severity scores (0-9). The system enables both onsite and remote access, automatic data logging, real-time image capture with PyQt5-based GUI, and environmental parameter control (temperature: 5-50°C, humidity: 40-90%RH, light: 0-15,000 lux), which can be manually set and automatically set based on the requirements of the user. This integrated approach demonstrates practical deployment of edge AI and IoT technologies for precision agriculture and disease monitoring applications.