Soybean is an important food source that is frequently preferred in animal feeds with its high protein value. However, soybeans contain many bioactive compounds that are antinutritional and/or poisonous. Urease is one of the most important of these. Processes such as extrusion is used to reduce these components' effect. Here, factors such as steam pressure and temperature affect the cooking level of the product. In the case of undercooked soybeans, components that harm animal health preserve their effect, while their nutritional value decreases in case of overcooking. The urease test has been used for many years to evaluate the cooking level of soybean. Here, according to the color change on the product as a result of the test, the cooking level is evaluated by an expert. This process is mostly done manually and is dependent on expert judgment. In this study, a machine learning-based approach has been proposed to evaluate the images of urease test results. Accordingly, samples were taken from the extruder during the processing of full-fat soybean. A data set consisting of over-cooked, well-cooked and undercooked sample images was prepared by performing the urease test. A binary classification process as cooked and undercooked and a classification process with three classes was carried out with four different machine learning models on the data set. In this way, it is aimed to both automate the process and minimize the problems that may arise from expert errors. Classification achievements of 96.57% and 90.29% were achieved, respectively, for two and three class tests with the CNN-LSTM model in 10-fold cross-validation tests.