A2BDefects

We present a high-resolution dataset tailored for automated defect analysis of internal load-bearing bricks in ancient architectures—a critical step toward integrating machine learning into cultural heritage preservation. Despite the increasing interest in applying computational methods to conservation, current datasets largely focus on external architectural features and coarse annotations, overlooking the intricate, safety-critical deterioration patterns hidden within internal structures. In collaboration with preservation experts, our dataset spans ten conservation zones covering architectural ensembles from the 11th to the 19th centuries.