Interplanetary Generalization and Adaptation of Deep Stereo on Mars

This dataset contains stereo navigation camera (NAVCAM) imagery and depth estimates from NASA's Perseverance rover of the Mars 2020 mission. It is curated as part of the MEP MAAARS project, aiming to improve stereo vision algorithms on Mars rovers using deep learning approaches. The primary purpose is to enable benchmarking of deep stereo algorithms against the flight-validated onboard stereo system under various illumination and terrain conditions on Mars. The dataset includes 16,768 synchronized and rectified stereo image pairs, camera parameters, along with corresponding onboard stereo's depth maps and FoundationStereo's depth estimates.