Application of Convolutional Neural Networks in the Classifying Growth Stages in Rice

Abstract

Rice is a staple food in many countries, making its efficient production critical to meet growing demand. This study introduces a novel approach by developing a machine learning-based classification system that accurately identifies the growth stages of rice crops—specifically, the "Vegetative," "Reproductive," and "Ripening" stages—using advanced convolutional neural network (CNN) architectures. Utilizing a unique dataset sourced from Roboflow, which includes annotated rice plant images, we meticulously divided the data into training, validation, and testing subsets to ensure robust model performance. Through the application of transfer learning on the ImageNet dataset, we explored the effectiveness of models such as ResNet50, InceptionV3, and MobileNetV2. Our findings indicate that InceptionV3 significantly outperformed the others, achieving a classification accuracy of 95.1% with a log loss of 0.13, compared to 87.3% and 93.5% for ResNet50 and MobileNetV2, respectively. This research not only demonstrates the potential of CNNs in precision agriculture but also provides practical insights into optimal model selection and data preparation techniques. This study highlights the potential of CNNs in precision agriculture and emphasizes the importance of model selection and data preparation in developing efficient crop monitoring and classification systems.