Automatic Classification of Mimbres Pottery Styles through Convolutional Neural Networks

Summary

This is an abstract from the "Research Hot Off the Trowel in the Upper Gila and Mimbres Areas" session, at the 86th annual meeting of the Society for American Archaeology.

This poster describes our attempt to address some long-standing questions about Mimbres pottery through convolutional neural network-based classifiers. Over the past few years the field of computer vision has made major strides in classification and segmentation tasks particularly due to the availability of rich training data and development of deep learning-based methods. We used over 8,000 images of Mimbres bowls to train a neural network to distinguish and sort bowls into Style I, Style II, and Style III types. After this initial training proved successful, we retrained the neural network to distinguish between micro-seriation types (e.g., early/late Style II and early/middle/late Style III) and bowls with design elements characteristic of the upper Gila or Mimbres River valleys. In addition, after the AI sorted the bowls into our specified categories, we used metadata associated with the bowls to address questions about production locales and changes through time in Mimbres pottery. Here we present the model we have developed along with results of its application on our dataset and also introduce a new web-based portal for upload of a new bowl for automatic classification into different styles.

This Resource is Part of the Following Collections

• Society for American Archaeology 86th Annual Meeting, Online (2021)   •
• Research Hot Off the Trowel in the Upper Gila and Mimbres Areas

Keywords

Spatial Coverage

Record Identifiers

Abstract Id(s): 32612