Corn is one of the most important crops in the United States and around the world. Farmers and agricultural managers need accurate ways to predict corn yield early in the season to make better decisions about irrigation, fertilization, and harvest. In this study, we tested how well drones equipped with different types of cameras can help predict corn yield in fields that were well-watered and fields that experienced water stress. We used visible light images (like regular photos), multispectral images (which capture more details about plant health), and thermal images (which detect heat) collected during different stages of the growing season to drive three machine learning models to see which methods gave the best results. We found that multispectral and thermal images worked better than standard photos for predicting yield, especially in fields with water stress. The Random Forest and Gradient Boosting models gave the highest accuracy. Importantly, we showed that farmers could get reliable yield predictions as early as the flowering stage, which can help them respond to crop needs in time. This research shows that combining drone images with advanced data analysis can improve yield prediction and support more efficient, precise farming.