When I bring a site model into a technical pipeline, I want two things at once: a useful visual answer and a record of how that answer was produced. PlaceMaker and Shapezo can both provide an early 3D context, but they expose different parts of the chain. PlaceMaker imports geographic data into an editable modeling environment. Shapezo takes a map selection and generates an initial AI-based model of the selected region.
This guide describes the workflow I use to keep those outputs comparable, inspectable, and clearly labeled.
1. Define the state before generating geometry
I create a small project manifest with the site boundary, coordinate reference system, units, north direction, ground datum, desired level of detail, and review question. I also record confidence for each input. Survey points may be high confidence. A public building height may be estimated. A Shapezo-generated roof is generated geometry until checked.
The manifest prevents a common failure: comparing two scenes that are actually in different units or orientations. It also gives me a place to store source dates and licensing notes for imagery, terrain, and map layers.
2. Build a controlled PlaceMaker context
In a SketchUp workflow, I use PlaceMaker to set the location and import only the layers needed for the question. That may include terrain, roads, building footprints, water, green space, and an image base. I keep each category in a named group and remove distant geometry that does not affect the study.
I treat the imported buildings as simplified city context, not detailed BIM. Before exporting, I check for missing heights, odd footprints, disconnected roads, excessive texture size, and a model origin that is too far from zero. Critical neighbors are replaced or corrected from better information.
3. Generate a Shapezo study scene
I draw a boundary in Shapezo that includes the project parcel and the surrounding features that influence it. I save the boundary itself, the map date, the orientation, and the generated output identifier. The AI scene is useful for quick massing and neighborhood discussion, but it is not a survey, permit model, engineering model, or guaranteed GIS dataset.
My first validation pass is visual and simple: Are the main streets continuous? Are the relative building heights plausible? Is the open space connected? Are steep grades, rail lines, or water edges visible? Any answer that depends on exact dimensions moves to a source that can support those dimensions.
4. Normalize before comparing
I never drop one scene directly on top of the other. I normalize units, origin, ground elevation, and north direction first. The transform is saved as a small text record or script so another person can reproduce it. The original exports remain untouched.
For each option, I use the same aerial, oblique, and eye-level cameras. Fixed views reveal whether a new camera is hiding a weak frontage or an incorrect grade. I compare only the geometry relevant to the question and keep background differences out of the design decision.
5. Track provenance at object level
I label major objects as observed, imported, generated, estimated, manually edited, or rebuilt. In a mixed scene, this matters more than a generic “AI-assisted” note. A reviewer can then ask why a building is generated, which layer supplied a road, or where a corrected terrain edge came from.
I keep prompts, boundary files, source dates, scripts, and screenshots beside the model. When a model changes, I record the change and the reason. A versioned folder structure such as source, generated, aligned, and reviewed is enough for a small study.
6. Add review gates before handoff
My review gates cover coordinate consistency, bounding-box sanity, polygon or face density, normals, material slots, and the behavior of any parametric edits. I also check that a map boundary has not moved during cleanup. A rendering pass is not a geometry pass; a nice image cannot certify topology, engineering standards, or final acceptance.
If the site supports a serious design decision, I rebuild the responsible geometry in the project-standard tool. Civil 3D may be appropriate for engineered surfaces, alignments, corridors, pipe networks, and quantities. Revit or another BIM tool may be appropriate for coordinated building information. Blender may be appropriate for mesh cleanup and rendering. PlaceMaker and Shapezo remain context and exploration inputs.
7. Choose by failure cost
I choose PlaceMaker when a wrong or missing geographic feature could change the conclusion and I need layer control or a traceable source. I choose Shapezo when the cost of waiting is higher than the cost of correcting an approximate first scene. The two can be chained, but the handoff should preserve provenance and uncertainty.
That is the reproducible version of a map-to-3D workflow: define the state, generate a scene, normalize it, label the geometry, test the relevant features, and rebuild what carries consequences. Speed is useful only when I can still explain what the model means.
Top comments (0)