NASA's Mars rovers have take and sent back to Earth thousands of images of rocks from the surface of the red planet. Some of these images feature small bumps on the rocks, called nodules. These bumps are interesting to the scientific community because they contain information about the history of water on Mars. Because there are so many images, and a single image can contain many nodules, it is very time-consuming to count and measure the nodules by hand. We trained a machine learning model to detect the nodules. To do this, we used 100 publicly available images, half of which featured nodules. The model takes in an image and produces a new image with just two pixel colors: white pixels where it 'thinks' there is a nodule, and black pixels otherwise. Using the model output, we can more quickly count and measure nodules on Mars.