Philippine Lime (Calamansi) Disease Detection and Classification Using YOLOv8 Model

Calamansi is one of the primary fruit crops cultivated in the Philippines. It is renowned for its acidic juice and offers numerous health benefits. It experienced a production decline of 1.5% to 12.903 thousand metric tons in early 2022, with the prevalence of calamansi diseases as one of the main factors. Calamansi farmers often neglect manual inspection due to its tedious nature, prompting the need for a more advanced approach. To address this, the researchers implemented transfer learning using YOLOv8 architecture to detect and classify calamansi diseases on the fruit, leaves, and branches. Eight dataset iterations using different labeling and augmentation methods were generated. The first four iterations included ten categories, comprising one healthy and nine disease classes such as Scales Infestation, Psorosis, Xyloporosis, Citrus Scab, Citrus Canker, Mites Infestation, Citrus Greening, Tristeza, and Powdery Mildew, and the next three iterations merged insect-related diseases, resulting in nine classes. Additionally, a final iteration of the dataset was generated as a fine-tuning technique to improve the model accuracy further. The model performance was evaluated across all eight dataset versions where the eighth version yielded the highest mAP@50 value of 87.2%.