This record is evidence-based and time-stamped. Sixteen frames were captured across two sides under photometric stereo. Sub-scores and PSA/BGS/CGC translations below are placeholder values until the Lumina scoring model is wired in.
Lumina's score maps to a probability distribution of grades for each major grader. Tap a tab to see the distribution. The dashed line marks this card's position on the Lumina axis; where it crosses a peak is the most likely outcome.
Capture. The card is scanned under eight programmable LED positions per side, sixteen raw frames total. The card never moves during capture; a single capture pass takes under 90 seconds for both sides.
Photometric stereo. The eight frames per side are fused into a composite that reveals surface topology — scratches, dents, print artifacts — by analyzing how shadows and reflections shift across the lighting set. This is what single-image photography misses and what makes the 3D viewer above possible.
Sub-score model. Vision models output sub-scores for surface, edges, corners, and centering using the composite + raw frames. Sub-scores are normalized against the Certified Raw population — not graded against a Platonic ideal, never rounded up.
Identity attestation. A separate model extracts a card-fingerprint embedding asynchronously. This is what lets Lumina recognize the same physical card across re-scans, registries, and time.
Models active on this report: Photometric Stereo v3 · Composite-Surface ML v4 · Identity Attestation v2 · Delta-Detection ML v1.