Topographic modeling using orbital imagery is a cornerstone of planetary photogrammetry and remote sensing, underpinning scientific exploration and analysis. While classical methods like stereo-photogrammetry (SPG) and (stereo)-photoclinometry (SPC) have long been developed, deep learning (DL) techniques have recently emerged as powerful alternatives, advancing rapidly in planetary topographic applications. This study briefly reviews the evolution of DL methods, contrasting their innovative approaches with the principles of traditional SPG and SPC techniques. We assess the efficacy of two representative DL models in reconstructing high-resolution topography for a large planetary body (the Moon) and a small asteroid (Itokawa), respectively. Our findings reveal that these DL methods successfully recover detailed terrain surfaces, even with limited input imagery, and produce results consistent with SPG- and SPC-derived models. These outcomes underscore the transformative potential of DL for efficient, robust topographic modeling across diverse planetary scales.