View Video Presentation: https://doi.org/10.2514/6.2021-4210.vidThis paper outlines the processes used during the development of a high-volume, high-fidelity lunar analogue dataset. Whether for terrain classification, novelty detection, or any other task, the need for such a dataset is significant. With investment in lunar exploration growing there is more potential for leverage image data than ever before; currently, such data is limited. The dataset described herein contains over 5000 images of an analogue lunar surface as seen from a front-facing rover camera. It also has pixel-wise terrain class labels and bounding box novelty labels, equipping it for supervised and semi-supervised training regimes. Beyond the dataset, the lunar analogue yard has also found application in remote operations testing and training of highly qualified personnel.