Rice bran is often adulterated with milled rice husks which can lead to losses and decreased production of livestock. Testing of rice bran adulteration is still carried out manually which is quite subjective, and thus it is necessary to test using the image analysis method based on convolutional neural network (CNN) through a visual image analysis process. The purpose of this study was to evaluate the accuracy of training, validation, and test data on rice bran adulteration with rich husks using the CNN-based image analysis. This study employed staining treatment using phloroglucinol. The stages of this research consisted of the process of mixing forgery, colouring treatment, taking pictures, sharing data, and building the CNN model. This study found that the results of the accuracy of training data, validation data, and test data with phloroglucinol staining treatment had accuracy that was still far from 100% accuracy. This result is due to the epoch value, quantity of datasets, and the different viewing angles when capturing an image.