Synthetic aged images improve cultural heritage retrieval when old photos are scarce
Compensating for Scarce Historical Images in Cross-Domain Cultural Heritage Retrieval Using Synthetic Aging
Computer Vision and Pattern Recognition
Summary
Matching photos of cultural objects taken long ago with recent pictures is hard because old photos look very different and there are often few of them. The authors explored making new 'fake old' images by adding signs of aging and damage to recent photos to help train computer models. They found that while these synthetic images don't fully replace real old photos, they can help improve recognition when few real old images are available. The benefit is strongest when only 25% to 50% of real old photos are present and gets smaller as more real old images are included.
cultural heritageimage retrievalsynthetic agingcross-domain recognitionEfficientNetV2training data scarcityimage degradationinstance-level retrievaldomain adaptation
Authors
Marcin Iwanowski, Adam Mazgaj, Ferdynand Gorski, Sabina Szymoniak
Abstract
Cultural heritage collections often contain contemporary and historical visual records of the same physical object. Linking these records is difficult because corresponding images may differ in viewpoint, acquisition conditions, color reproduction, framing, resolution, and degradation, while genuine historical images are frequently scarce. This study investigates whether synthetically aged contemporary images can replace or complement missing historical training data in bidirectional instance-level retrieval. Synthetic old-domain images are generated using degradation-oriented transformations. An EfficientNetV2-M model is evaluated on identity-disjoint training, validation, and test sets across three dataset partitions and three training seeds. Mixed real-synthetic training is compared with real-only baselines using proportionally scaled and fixed 300-batch-per-epoch schedules. Complete replacement of genuine historical images reduced bidirectional mean R@1 from 86.56% to 81.27%, showing that synthetic aging does not reproduce the full genuine old-domain variability. Increasing the number of independently generated synthetic variants provided no consistent improvement. Under controlled scarcity, however, synthetic completion improved mean R@1 by 3.69 percentage points at 25% genuine historical coverage and by 2.92 points at 50%, relative to the proportionally scaled real-only baselines. At 75%, the gain decreased to 2.00 points, while performance remained comparable to the complete-real-data reference. Fixed-schedule real-only controls did not reproduce these improvements. The results indicate that genuine and synthetic observations are complementary. Synthetic completion primarily benefits retrieval by extending cross-domain identity coverage rather than by increasing training exposure, with its contribution gradually decreasing as genuine historical coverage increases.