The dataset consists of 6,691 images collected by three instruments of the Mars Science Laboratory (MSL, Mars Science Laboratory) (right eye Mastcam, left eye Mastcam and MAHLI), covering 24 categories. These images are "browsed" versions of each raw data product, not full resolution, each image is approximately 256×256 pixels.

The researchers divided the MSL images into training, validation, and test sets based on their acquisition date (Martian day, or sol). This strategy was chosen to simulate how the system would be used operationally as the image archive grows over time. Images were acquired from sol 3 to sol 1,060 (August 2012 to July 2015). The exact train/validation/test split is given in separate files.