Needles in the Landscape: Semi-Supervised Pseudolabeling for Archaeological Site Discovery under Label Scarcity

In three linesSemi-supervised semantic segmentation model for predicting undiscovered archaeological site locations. Uses dynamic pseudolabeling and CRF-RNN to handle severe label scarcity. Matches LAMAP performance on DEM data, improves Dice scores on raw satellite imagery.

Summary generated by Claude — human-verified