The challenge
The client needed segmented 3D meshes reconstructed from images, with each object identified, for use in analysis and downstream tools.
What we built
An image segmentation model combined with adaptive point cloud generation and Poisson surface reconstruction.
What we delivered
- Segmentation and classification of objects in images
- Point clouds generated for each segment
- Adaptive resolution of the point cloud
- Poisson reconstruction to create the surface mesh, with Open3D
- Semantic classification of the 3D mesh objects
The outcome
Semantic segmentation of 3D objects and geometrically accurate meshes.
