Abstract
In this paper, we discuss the problem of object detection in a cultural heritage application context. In particular, the objects to be detected are ancient rock carvings, discovered at the archaeological site of Vathy, Astypalaia in Greece. Without the help of a marker or a human expert, the rock carvings are extremely difficult for a visitor of the site to discern from their surroundings. We explore the possibility of using a computational method that could replace the human expert and detect the rock carvings of interest without the aid of a specific marker. We present a dataset of images that is comprised of annotated photographs of the rock carvings, taken in situ and under differing poses and lighting parameters. Two methods for detection are applied; the first method makes use of a supervised, deep learning-based model, while the other relies on feature point-based matching to an annotated template, in the context of which we propose a simple image matching distance. We show that each method is applicable under different conditions, and evaluate their effectiveness with numerical trials.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have