A correlation between thin-film nitrate sensor performance and sensor surface texture was hypothesized. Based on this hypothesis, we began research on the application of machine learning methods on thin-film nitrate sensor surface images to predict its performance. This technology would enable real-time optimization adjustments to be made during production to greatly increase the quality of the sensors while reducing costs associated with testing and defective sensors. Recently, we have made progress in the addition of new texture features, repeated crossvalidation methods, and auto-tuning of hyperparameters.