Convolutional Neural Networks (CNN) are a class of deep neural networks widely applied to analyzing images. Several authors have used this method to successfully classify microfossils, core images, petrographic photos, and rock and mineral hand sample images in the geosciences. This work hypothesizes that MobileNet (a pre-trained convolutional neural network) is helpful in classifying sedimentary structures and reducing interpretation bias. We used 637 images of cross-lamination, planar lamination, and ripples taken in the field from Nasa galleries and Google images to train and test the CNN. The expert defined the labels (type of sedimentary structure) needed to train the algorithm and also monitored the result to address incorrect classification. The accuracy achieved in predicting sedimentary structures on new images using MobilNet was 82%. This tool could help identify sedimentary structures in the field without regarding the geologist’s level of expertise or prior experience.