Artificial Intelligence Based Web tool for Classification of Rice Panicle Stages

Rice stands as a fundamental global staple, playing a pivotal role in ensuring food security worldwide. Accurate identification of key rice growth stages, including booting, heading, anthesis, grain filling, and grain maturity, holds paramount importance for agricultural decision-making. Nonetheless, a notable gap persists in leveraging red-green-blue (RGB) images for stage recognition. This research leverages cuttingedge computer vision and deep learning techniques, specifically the EfficientNetB0 convolutional neural network algorithm, to bridge this gap effectively. Demonstrating exceptional performance, EfficientNetB0 achieves an impressive overall accuracy of 82.8 percent. A granular examination of growth stages unveils varying accuracy levels, with boot leaf emerging as the most reliably detected stage at 95.1 percent, while anthesis poses the greatest challenge at 72.28 percent. Web tool based on developed EfficientNetB0 model to classify rice panicle stage has been developed using Flask framework of Python. This study marks a significant stride in automated monitoring capabilities, empowering researchers with timely insights for informed decision-making in agriculture.