Photo montage: the design in a photograph of the plot

What it gives you

• The real surroundings, not invented ones

  Street, neighbours, trees and light come from the photograph. The image model may only render the building; outside the mask every pixel of the photo stays.

• A camera from point pairs, not from feel

  Four pairs fix position and direction, six the focal length as well. The error per pair shows whether the solution holds.

• Before and after on one page

  The share page shows photo and montage in one wipe. Clients and neighbours see what changes – and what does not.

• Remove what goes

  An old building or shed that will be demolished can be marked as an area; the image model replaces it with the continued surroundings.

How it works

1. 01Upload the photoIn the “Photo montage” tab choose a JPG, PNG or WebP, ideally taken from the pavement or the neighbouring plot with the horizon in the frame. The browser shrinks the file to 2000 pixels; date and focal length are taken from the EXIF data.
2. 02Align the model roughlyTurn and zoom with the mouse, pick the angle of view with the focal length slider (18 to 85 mm) and drag the dashed line onto the horizon of the photo. The model lies over the photo, half transparent with dark edges.
3. 03Set point pairsTool “Point in model”: click a corner of the house, it snaps to the nearest edge. Then “Point in photo”: click the same corner in the photograph. Four to eight pairs on different edges, not all on one line.
4. 04Solve the camera“Solve camera” computes position, direction and tilt, from six pairs the focal length too, and shows the error of every pair in pixels. A pair with a large error is usually misplaced: remove it and click again.
5. 05Remove existing structuresIf something should disappear from the photo, click a polygon over it with “Remove existing” and close the area. It joins the mask.
6. 06Produce the pictureChoose the quality level and “Produce picture”. The result opens as a before/after slider; on the share page it stands under “On the plot”, with download and remarks like any picture.

What the browser does and what the image model does

The camera is solved in the browser without a model call: a small least-squares fit over position, rotation, tilt and angle of view, with roll fixed at zero because a photo of a plot is held level. From the solved viewpoint three pictures are made in the photo’s aspect ratio: the photo with the grey model and its shadow, the model’s white silhouette on black, and the depth map.

These three go to the image model with the instruction to render inside the silhouette only and to take light, colour, sharpness and grain from the photograph. Whether the image model keeps the boundary exactly cannot be promised; the instruction is strict, and the comparison with the original shows deviations at once.

Photographs that work well

Pictures at eye level with a visible horizon, without strong wide-angle distortion and without zoom beyond 85 mm. Corners that are sharp in the photo: plot boundaries, neighbouring buildings, fences, kerbs. The wider the pairs are spread across the frame, the more stable the solution.

Blur people and number plates before uploading: the photo goes to the image model and is shown on the share page. What we process is described in the privacy policy under “Fotos des Grundstücks”.