NASA asks citizens to analyze Mars photos to improve rover algorithm

NASA wants future Mars rovers to be able to recognize the surface of Mars and, for example, rock formations, and is calling for the help of civilians. The AI4Mars program allows people to view images of Perseverance and label certain rock formations.

AI4Mars has been around since June 2020 and allowed people to circle stones and sand plains in images taken by the Curiosity Mars rover, for example. NASA has now expanded this program to include images from Perseverance and additional labels. For example, the extended program allows citizens to label float rocks, or ‘islands of rocks’, and nodules. The latter are pebbles the size of air rifle bullets made by water and compressed minerals.

Users can now circle and label certain surface features within AI4Mars. So far, citizens have tagged nearly half a million photos. The tagged photos improve NASA’s SPOC AI algorithm. SPOC stands for Soil Property and Object Classification and identifies surface features on photos sent by Mars rovers. Now SPOC would have an accuracy of almost 98 percent.

Perseverance sends tens to hundreds of photos to Earth every day; too many photos for scientists to sift through on a daily basis. Especially since scientists decide on the basis of these camera images where the Mars rover should drive, it is important that these images are analysed.

Algorithms such as SPOC could therefore help future Mars rovers to analyze photos themselves and determine which ones should return to Earth. If necessary, SPOC should give Mars Rovers more self-driving functions. SPOC could also analyze archival photos of Mars rovers to provide scientists with information about the Martian surface more quickly. AI4Mars is available on the Zooniverse site.