Shape-from-x is a fundamental class of problems and associated methodologies within computer vision focused on the inference of three-dimensional scene or object geometry from two-dimensional image data. This research area investigates how diverse visual cues ('x'), such as shading (shape-from-shading), texture (shape-from-texture), multiple views (shape-from-stereo), or image sequences (shape-from-motion), can be leveraged to recover properties including depth, surface normals, and full 3D models, serving as a cornerstone for 3D reconstruction, scene understanding, and spatial perception from visual input.