AI uses changing light to group fragments of Viking textiles

The Oseberg tapestries survive as fragments showing people, weapons, animals, carriages and houses. The Museum of Cultural History’s TexRec project is investigating more than 80 pieces from several tapestries, asking how their scenes once formed narratives. Their fragility makes unnecessary handling a problem.[1]

In a study published in February, researchers photographed textile samples under 50 lighting directions, then fed a derived reflectance image into a pretrained ResNet-50 network. Its visual features supplied the basis for grouping similar fragments.[3]

Keep the cloth still, change the light

Reflectance Transformation Imaging, or RTI, begins with a stationary camera and a sequence of differently illuminated photographs. The changing highlights and shadows carry information about the surface. Software fits a representation that can be relit on screen, allowing a researcher to inspect fine details without repeatedly repositioning the object.[2]

Cultural Heritage Imaging describes the resulting file as an ordinary-looking image whose pixels retain information about how they respond to light. Moving the virtual light changes what becomes visible. That makes the illumination part of the measurement, rather than a photographic setting chosen once and forgotten.[2]

The textile study used the model’s base-reflectance component. One RTI grouping placed Oseberg fragments 10, 12 and 16 together, consistent with an archaeological hypothesis. A separate Polish textile supplied control images with known origins.[3]

A suggestion to examine

The silhouette score, which measures cluster separation and compactness, favoured ordinary photography: 0.425 versus RTI’s 0.404. A second measure of cluster separation, the Calinski–Harabasz score, favoured RTI at 18.5 versus 15.6; higher is better for both. Similarity groups remain suggestions for investigation, with physical joins unresolved.[3]

TexRec’s wider work combines imaging with thread and weave analysis, dye chemistry and conservation research. Much of the original colour has faded; identifying surviving breakdown products may help reconstruct it. Those questions require evidence from the material as well as an arrangement of images.[1]

Grouping fragments changes where an archaeologist looks next. A convincing neighbour on screen becomes a candidate for comparison: does its weave agree, does the surviving scene make sense, and is there a material connection? The value is in making that search more directed while leaving the original cloth undisturbed.

Report an error or suggest a correction