Related Experiment Video
Updated: Sep 2, 2025
13:19
Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
9.3K
Machine learning for stone artifact identification: Distinguishing worked stone artifacts from natural clasts using
Joshua Emmitt1, Sina Masoud-Ansari2, Rebecca Phillipps1
1School of Social Sciences, University of Auckland, Auckland, New Zealand.
Plos One
|August 10, 2022
Summary
Machine learning technology accurately identifies stone artifacts from natural rocks, aiding archaeologists. This AI tool surpasses human analyst performance in classifying worked stone objects.
Area of Science:
- Archaeology
- Computer Science
- Artificial Intelligence
Background:
- Stone artifacts are crucial archaeological finds, yet their identification is bottlenecked by expert availability.
- Distinguishing worked stone objects from natural lithic clasts requires specialized analytical skills.
Purpose of the Study:
- To develop and evaluate a machine learning (ML) based technology for automated stone artifact identification.
- To address the expert analyst shortage in archaeological lithic analysis.
Main Methods:
- A dataset of 6769 2D images (3868 artifacts, 2901 rocks) from Egypt, Australia, and New Zealand was compiled.
- A machine learning model using PyTorch's Faster R-CNN ResNet 50 implementation was trained and tested on the image dataset.
Main Results:
- The ML model achieved 100% agreement with original human classifications.
- The model's performance exceeded that of two independent human analysts reassessing the same images.
Conclusions:
- Machine learning offers a consistent and accurate solution for identifying stone artifacts in large archaeological assemblages.
- This technology has the potential to significantly enhance archaeological research by automating lithic analysis.
