End-To-End Deep Learning to Assess Archaeological Potential from Multimodal Data

We propose a novel deep learning-based workflow to predict the Archaeological potential in the hinterland of ancient Sagalassos in South- west Turkey. Our approach concerns a convolutional architecture trained on a semantic segmentation task with self-supervised pretraining. Our experimental results demonstrate the potential for end-to-end learning techniques in assessing archaeological potential using multimodal data sources.