[servantftransperfect]

Description

Adds aliceVision_meshResection and its MeshResection Meshroom node. The tool estimates camera poses from 2D/3D correspondences that a user picks by hand between the images and a reference mesh.

• Input: an SfMData file and a JSON shape file. The JSON holds a list of SurveyPoint shapes, keyed by viewId. Each observation has its 3D position (X, Y, Z), its 2D position (x, y) and two flags, picked2d and picked3d.
• Processing: for each view, the tool uses only the observations where both picked2d and picked3d are true. It runs resection on them with SfMLocalizer and skips views that have too few correspondences.
• Output: an SfMData file with the poses it estimated. It also writes the reprojection residuals (ex/ey) back into the JSON shape file.

Implementation remarks

• The node gets its observations from desc.ShapeList / desc.SurveyPoint and writes them to JSON in preprocess, so it needs a Meshroom version that supports these shape attributes.

🤖 Generated with Claude Code