/ OMAR ATABANY
Lighting and reconstruction gaps at NMEC
A closing capture session at the National Museum of Egyptian Civilization, and why some of that batch reconstructed with gaps photogrammetry usually shouldn't leave.
Just closed out the last capture session at the National Museum of Egyptian Civilization (NMEC). That's the final batch of artifacts from this location: a full round of images collected across the gallery, same walk-around-the-case method as everywhere else. Not every piece from it is coming out clean.
Photogrammetry reconstructs geometry by matching the same surface feature across overlapping frames shot from different angles. That only works if the surface looks consistent as the camera moves: same shading, same highlights, roughly the same appearance from frame to frame. NMEC's gallery lighting is built for viewing, not capture: tight spotlighting per case, strong directional contrast, in some cases a light source close enough that its highlight shifts across the object's surface as the shooting angle changes rather than staying fixed to the object itself. A moving highlight isn't a consistent feature. From one frame to the next it can look like a different point on the surface entirely, and the matcher just fails to connect it to anything.
The result isn't a texture problem, it's a geometry problem: sections of the mesh where matching failed outright end up as literal holes, not filled with the wrong thing but not reconstructed at all. That's different from a case like Weeping Lady's unseen back, where a smooth alpha-wrap patch stands in for geometry that was never going to be photographed in the first place. Here the surface was in frame; the lighting just made it unusable for the reconstruction.
Nothing about this changes the plan for future NMEC visits, since there aren't going to be more. This batch is what there is. For the pieces with gaps, the options are the same alpha-wrap patching used elsewhere, or reprocessing with a tighter subset of frames that avoids the worst of the highlight sweep, if there's enough clean coverage left to rebuild from. Until one of those happens, they stay flagged as not release-ready rather than shipped with holes in them.
Archive note, 14 October 2026: this is the session report as written on 19 August. Its release-status statements describe that stage of the work; consult each current object record for today’s downloads, processing and remaining limitations. The lighting explanation is the project’s working diagnosis, not an independently isolated cause.