A Deep Learning Approach to Classify Shell Rings

Bring the serenity of the seashore into your everyday style with this glistening seashell ring. Embodying the beauty of sand and sea, this delicate ring symbolizes femininity and new beginnings. Seashells have also long been a symbol of transformation, representing the mysteries and treasures that are uncovered by the ebb and flow of ocean life.

The mounded deposits known as shell rings or middens are a feature of Georgia’s coastal landscape and are well-known in Southeastern archaeology. Located along barrier islands and coastal mainland, these circular to semi-circular deposits of shell, bone, soil, and artifacts represent the remains of the food of native groups that occupied these areas in the Late Archaic Period (4,200–3,000 years ago). Archaeological investigations into middens on St. Simons and Sapelo Island suggest that these features were a result of household deposition rather than ceremonial mounding, and may have begun as households before being mounded.

Shell rings also contain information about past climates. Similar to tree rings, molluscs develop increments in their shells that record the conditions during which they grew. This enables scientists to reconstruct the environmental conditions of a particular location in the same way that we can calculate a tree’s age by counting its rings.

A deep learning model was trained using multispectral Sentinel-2 data to detect and classify shell rings. Using the Mask R-CNN approach, the model achieved 75% validation accuracy. Multispectral evaluations of the validation set revealed distinct differences between confirmed and suspected archaeological shell rings and the surrounding landscape, with greater clarity in the single-polarization bands. While these results are promising, further evaluation of this method is necessary to minimize false positives. seashell ring

Leave a Reply

Your email address will not be published. Required fields are marked *

Back To Top