Depth Estimation

Monocular depth estimation uses neural networks trained on paired image-depth data to predict a depth map from a single photo. The model learns visual cues that humans use for depth perception, including relative size, occlusion, texture gradient, atmospheric haze, and perspective convergence. The resulting depth map assigns a distance value to each pixel, enabling 3D-aware editing operations. Common uses include synthetic bokeh, parallax effects for social media, AR object placement, and better AI segmentation. Accuracy can drop around hair, glass, reflective objects, flat textures, and unusual camera angles, so depth-aware edits usually need a quick edge review before export. Smartphone portrait modes use depth estimation to separate subjects from backgrounds for real-time blur effects.