This research aims to estimate a plant’s water stress using advanced deep-learning techniques and stomata micrographs. Over 100 microscopic images of Oryza sativa (rice) were captured and annotated with open and closed stomata as part of the dataset. The images were captured under an optical lens with 40× zoom. Standard preprocessing steps such as image resizing and gray-scaling were performed to enhance model performance and reduce complexity. The YOLOv8 model was then trained over 600 epochs resulting in an accuracy of 92% in stomata detection and classification. A high ratio of open to closed stomata was observed in the test images, implying that the plant was not experiencing water stress and maintained a healthy state. The study can facilitate the currently manual and time-consuming process of stomata annotation. Further research is necessary to generalize the model for different stomata shapes. Integrating this technology into IoT networks can help reduce inefficiencies in water management in agriculture.