Autonomous image recognition has numerous potential applications in the field of planetary science and geology. During exploration, geologists could encounter an unknown rock and instead of having to bring back the sample to the laboratory for analysis, a better approach would be to have a mobile device classify the image of a rock. As well, instead of waiting a long time for a planetary rover to send back an image to Earth for classification, its on-board computer could have a software that could automatically classify images of outcrops. In 2017, Shu et. al. used a Support Vector Machine (SVM) classification algorithm to classify uniform rock images into 9 different classes with the image features extracted autonomously. Through this method, they achieved a classification accuracy of 96.71%. Recent publications have shown that Convolutional Neural Networks (CNNs) perform better than other algorithms in classifying images of everyday objects, more specifically for the ImageNet dataset. In light of this development, this paper demonstrates the use of CNNs to classify the same set of rock images. With the addition of dataset augmentation, a 3-layer CNN is shown to have a significant improvement over Shu et. al.’s results, achieving an average accuracy of 99.60% across 10 trials on the test set. Having proven that CNNs can classify uniform and clean images of rocks, this research then tackles a more interesting and practical problem in classifying natural scene images of rocks where the images are taken during field exploration without a standardized method and specialized equipment. The task has been simplified into a binary classification problem where the images are classified into breccia and non-breccia. This research shows that a 5-layer CNN achieves 89.43% classification accuracy for this task.