Projecting Geometries onto a Model

This example demonstrates how to project a geometry onto a trained GeomAI model and retrieve its latent parameters.

Before you begin

• Complete “Building a GeomAI Model” to train a GeomAI model.

• Ensure that the model training has completed successfully and that its workspace is available.

• Prepare a geometry file in VTP or STL format.

Import necessary libraries

import ansys.simai.core as asc
from ansys.simai.core.data.geomai.predictions import GeomAIPredictionConfiguration

Configure your settings

Update these variables with your specific settings:

ORGANIZATION = "my_organization"  # Replace with your organization name
WORKSPACE_NAME = "new-bracket-project #1"  # Replace with your trained model workspace
GEOMETRY_FILE = "./td/bracket_264.vtp"  # VTP or STL file to project

Initialize the client and get the workspace

simai_client = asc.SimAIClient(organization=ORGANIZATION)
geomai_client = simai_client.geomai
workspace = geomai_client.workspaces.get(name=WORKSPACE_NAME)
print(f"Using workspace: {workspace.name}")

Upload a geometry

print("Uploading geometry...")
geometry = geomai_client.geometries.upload(GEOMETRY_FILE)
geometry.wait()
print(f"Uploaded {geometry.name} with id {geometry.id}")

Project the geometry onto the workspace model

The projection operation is idempotent: if this geometry has already been projected onto the model, the existing projection is returned.

print("Running projection...")
projection = geometry.run_projection(workspace)
projection.wait()

latent_parameters = projection.get_latent_parameters()
print(f"Projection {projection.id} on model {projection.workspace_id}")
print(f"Retrieved {len(latent_parameters)} latent parameters")

Reconstruct the geometry from its latent parameters

The latent parameters can be passed to a prediction configuration to regenerate the geometry from the model.

config = GeomAIPredictionConfiguration(
    latent_params=latent_parameters,
)
prediction = geomai_client.predictions.run(config, workspace)
print(f"Prediction {prediction.id} started...")
prediction.wait()

output_path = f"reconstructed_{prediction.id}.vtp"
geomai_client.predictions.download(prediction.id, output_path)
print(f"Saved reconstructed geometry to {output_path}")

Look up an existing projection

Retrieve a projection without starting a new computation. The result is None if the geometry has not been projected onto this workspace’s model.

print("\nLooking up an existing projection...")
existing_projection = geometry.get_projection(workspace)
if existing_projection is None:
    print("Geometry has not yet been projected onto this workspace's model.")
else:
    print(f"Existing projection: {existing_projection.id}")
    print(f"First three parameters: {existing_projection.get_latent_parameters(n=3)}")

List projections in the workspace

projections = workspace.list_projections()
print(f"\n{len(projections)} projections in {workspace.name}")
for proj in projections:
    print(f"- {proj.id}")

Next steps

To explore the model further, you can:

• Generate geometries from latent parameters.

• Interpolate between geometries: Interpolating Between Geometries.