Bones of Yew

Purpose-Built 3D Scan Data for AI Training

Bones of Yew is one of ours: a project where we scanned environments and subjects specifically to produce AI training data… not the scraped, unlicensed imagery that most models are still trained on.

We captured everything using terrestrial LiDAR and photogrammetry… point cloud processing first for the geometry, then high-resolution photogrammetry layered on top to deliver fully textured meshes with the geometric and material detail that training systems actually need. Photography gives you flat colour and tone. Scan data gives you spatial relationships, surface normals, measurable depth… the things a model needs to understand an object rather than just recognise its silhouette.

This is a part of the business we’re deliberately growing. As AI transparency and rights become a live issue across film and the wider creative industries, there’s a real difference between a dataset built from consented, purpose-captured scans and one assembled from other people’s uncredited work. We think that difference matters, and Bones of Yew is us putting that into practice rather than just talking about it.

Deliverables followed our usual workflow: registered point clouds, clean textured meshes, and documentation suitable for direct use in a machine learning pipeline, no reformatting required on the other end.