Rice fields based on Nitrogen levels using Learning Vision Transformer with Squeeze and Excitation

The main component of chlorophyll needed for photosynthesis is nitrogen. Adequate nitrogen absorption will result in faster and healthier rice plant growth. Adequate nitrogen will increase the yield and quality of rice grains. This study aims to determine the proper nitrogen levels so that using nitrogen fertilizers becomes more measurable and environmentally friendly. Drones monitor large land areas, and deep learning helps with pattern recognition. Previous techniques require ground truth to distinguish between different objects. Our system will first learn from the collected files to obtain an optimal model. Predictions will be made using the model as a guide. The application of the Learning Vision Transformer with Squeeze and Excitation is its contribution. This approach can sharpen attention to important feature channels by dynamically adjusting weights. The dataset consists of crop data collected from rice fields with four classes of nitrogen conditions. The average training accuracy is the result obtained. According to the results, the average model accuracy rate is 99.5%, and the training computation time is about 9 minutes. RMSE = 0.0685, MAE = 0.0047, and MSE = 0.0047. The confidence level of accuracy in identifying new data or experimental data varies from 93.5 to 99.8%.