An important aspect of characterizing exoplanets is getting reliable information about structures on their surfaces. Resolving exoplanet surfaces might seem like an impossible challenge. Given that no resolved direct images of exoplanet surfaces exist, this skepticism is understandable. Terrestrial exoplanets are small, distant, and difficult to observe due to the stark contrast with their host stars.The Small ExoLife Finder (SELF), a hybrid interferometric telescope currently under construction at Teide Observatory, Tenerife, is a dedicated instrument for direct imaging of exoplanets. It serves as a prototype for a much larger telescope, the ExoLife Finder (ELF). Both employ advanced optics and photonics, including lightweight mirrors and a novel approach to starlight suppression through destructive interference. With SELF, Jovian exoplanets will be observed, paving the way for ELF to target terrestrial ones.Despite these promising prospects, obtaining resolved surface maps of exoplanets from the surveys of these telescopes remains non-trivial. Even the most sophisticated telescopes planned to date will not be able to directly resolve exoplanet surfaces. This task requires an alternative approach: reconstructing exoplanet surfaces from unresolved reflected light curves. Applying deep learning to this inverse problem and testing this spin-orbit tomography approach on Earth as an exoplanet shows the robustness of this method at recovering compact structures on exoplanets such as continents, even at moderate signal-to-noise (SNR) conditions.Further characterizing, combining light curves across different wavelengths, even allows distinguishing between vegetated land, deserts, or ice. This would be an important contribution to the search for biosignatures. Going beyond natural features, this technique is also promising for discovering large-scale artificial structures, which is a highly interesting path to technosignature detection.