Integrating Deep Learning for Plant Growth Monitoring and Browning Disease Severity Identification in Coconut Tissue Culture

The research aimed to address the inefficiencies and inaccuracies associated with manual plant growth measurement methods at the Coconut Research Institute (CRI). Deep learning algorithms were employed to automate the analysis of plant growth factors, including leaf size, leaf count, root length, and stem growth. Images of coconut tissue culture plants were collected and processed, followed by data augmentation to enhance model accuracy. The YOLOv8 model was utilized for the detection of flasks and the segmentation of leaves and stems. Through iterative training and augmentation, models with improved precision, recall, and Mean Average Precision (mAP) metrics were developed. The model, trained on an augmented dataset with 50 epochs, demonstrated significant improvements in segmentation tasks, achieving a precision of 0.925 and mAP50 of 0.884. For flask detection, a smaller dataset and pre-trained weights led to the selection of a model with a precision of 0.990 and mAP50 of 0.995. Root classification was performed using a YOLOv8 model, with the highest accuracy observed in the sharpened dataset (0.6977). Additionally, four CNN models were tested for browning disease classification, with InceptionV3 achieving the highest accuracy of 96%. The study demonstrates that deep learning models, particularly YOLOv8 and InceptionV3, are effective in automating plant growth monitoring and disease detection, significantly improving the accuracy and efficiency of data collection at the Coconut Research Institute in Sri Lanka(CRISL). The results provide a foundation for the broader application of these models in the agricultural domain, enhancing overall plant health management.