Towards the Identification of Archaeological Elements Through Machine Learning: a New Possible Tool for Archaeological Survey

Abstract

This is an exploratory work that aims to promote an archaeological survey methodology based on computer vision to classify photographs of small archaeological objects. To achieve this, the system was taught to recognize archaeological remains through a sample of photographs of lithic debris and ceramic fragments found on the surface in prospecting campaigns in the Great North of Chile, and then asked the system to analyze and recognize another sample of the same types of objects. Preliminary results show an error rate of only 6.6% in the developed tests. This work opens the possibility of using this technique based on low-height photographs taken with drones, which would allow it to obtain more precise results, less expensive and in less time.