Shed Happens

Contractors, builders, and tradespeople measure things constantly — lumber, clearances, site dimensions — and they do it with tape measures, laser rangefinders, or rough guesses from photos. Shed Happens replaces all of that with a phone camera.

Point a phone at any object. The app captures a sequence of images, runs photogrammetry to reconstruct the 3D geometry, and uses AI-based scaling to map the model to real-world dimensions. The result is an accurately scaled, exportable 3D model that holds up in the field — measurable in-app and shareable as GLB, FBX, or OBJ for downstream workflows.

Built with a React front end for the UI layer and Unity for the 3D pipeline, Shed Happens ships cross-platform on iOS and Android. The Unity viewer embeds via WebGL inside the React shell, handling model display, measurement tools, and AR overlay while React manages user accounts, scan history, and export.

Mobile photogrammetry exists, but the existing tools are built for 3D artists, not construction crews.

Apps like Polycam and KIRI Engine produce good meshes, but they don't solve the scaling problem. Without ground control points or LiDAR, a photogrammetry model has no inherent sense of real-world size. A 3D scan of a shed could be 3 meters or 30 — the geometry is correct, but the scale is meaningless without a reference.

The client needed something a site foreman could use without training: point, shoot, get a measurable model. No markers, no calibration rigs, no post-processing on a desktop. The accuracy had to be good enough for construction estimation — within a few centimeters of tape-measure truth.

Feative Studios built a hybrid pipeline. The app collects overlapping images from the phone camera, sends them to a cloud backend running COLMAP for Structure-from-Motion reconstruction, and produces a dense point cloud that gets meshed and textured on the server.

For scaling, the system fuses multiple signals: EXIF GPS metadata constrains camera positions, known construction element sizes (standard door widths, lumber dimensions) serve as implicit scale references, and a lightweight AI depth estimation model provides metric depth per frame. The combination resolves the scale ambiguity without requiring the user to place physical markers.

The React front end handles scan management, user accounts, and the export pipeline. Unity renders the 3D model in-app with measurement tools — tap two points, get the real-world distance. AR mode overlays the model onto the physical environment for visual comparison. Export formats cover GLB for web/AR, FBX for professional CAD workflows, and OBJ for legacy compatibility.

Shed Happens delivers sub-10cm accuracy on typical construction objects when 50+ images are captured with reasonable overlap. For smaller objects with dense coverage, accuracy tightens to the low single-digit centimeters — comparable to consumer laser rangefinders at a fraction of the cost and weight.

The app runs on any modern iPhone or Android device. Scans process in under two minutes on the cloud backend, and the resulting models are immediately measurable in-app or exportable for use in estimation software, CAD tools, or AR presentations. The client uses it for site documentation, remote measurement, and client-facing 3D walkthroughs of proposed builds.