Many planetary surfaces are heavily cratered as they witnessed the early stages of Solar System evolution during which impact cratering was a frequent process. Upon impact, rock fragments are ejected from the crater cavity and deposited elsewhere on the surface, where they potentially form secondary craters. The unknown contribution of secondary craters increase crater density and distort crater statistics, which ultimately biases the estimated age of a surface unit, a key diagnostics for understanding the evolution of planetary bodies. The size and velocity distribution of the ejected rock fragments is a poorly understood aspect so that an important link between crater statistics and planetary surface age keeps missing. One way to close this connection is to make use of the population of boulders (meter-sized rocks) that can be detected on high-resolution images of planetary surfaces, such as the Moon’s. Boulders are the only remnants of the ejected materials and their size and shape as well as the terrain on which they are found provide important insight into the ejection mechanisms. BOULDERING aims to advance the detection of boulders on planetary surfaces from high-resolution imagery using deep learning and to compile size and shape distributions of boulder populations. Based on this, this project will boost our understanding of cratering records and the implications for planetary surface evolution. A versatile automatic boulder detection algorithm will be developed using a convolutional neural network. This algorithm will first be validated on terrestrial boulder populations in Death Valley and the Mojave Desert and will then be trained with remote sensing data for application on the lunar and martian surfaces. By following this approach, ground data collected on Earth will be used to test the algorithm’s capacity to measure the sizes and shapes of boulders, which is key to make robust inferences on the boulder population on other planetary bodies.