DepthScape: Authoring 2.5D Designs via Depth Estimation, Semantic Understanding, and Geometry Extraction

2.5D effects, such as occlusion and perspective foreshortening, enhance visual dynamics and realism by introducing 3D depth cues into 2D designs. However, creating these effects remains challenging, as designers must manually infer and author depth relationships—such as relative ordering, occlusion boundaries, and perspective scaling—within 2D representations. We introduce DepthScape, a human–AI collaborative system that facilitates 2.5D effect creation by placing design elements directly into 3D reconstructions. Using monocular depth reconstruction, DepthScape transforms images into 3D scenes, enabling depth-based blending that produces realistic occlusion and perspective foreshortening. To simplify 3D placement, DepthScape leverages a vision-language model to analyze source images and extract key visual components as parametric anchors, which support direct manipulation editing. The system design was iteratively refined through a formative user study with an early prototype. We evaluate DepthScape through a technical study on 100 professional stock images to assess robustness, alongside an expert evaluation confirming design quality, usefulness, and broad application potential, further illustrated through five example scenarios.