 This paper presents a novel framework based on deep learning, DL, for predicting soil moisture content, SMC, from synthetic aperture radar, SAR, images. To improve the accuracy of the predictions, a sparse autoencoder, DL network was used to reconstruct the available dataset. Additionally, a Bayesian optimization strategy was employed to optimize the hyperparameters of the ML models. The results of the study show that the use of DL augmented data can significantly improve the prediction performance of the 5 ML models compared to the original dataset. The Gaussian process regression, GPR, model achieved the lowest root mean squared error, RMSC, of 4.05%, while the random forest, RF, model had the highest R2 value of 0.81. This article was authored by Muhammad Dabour, Gata Atiyah, Soham Mishul, and others.