An open-source AI suite called PyPottery claims it can speed up pottery documentation by as much as 40 times.
Researchers built PyPottery as a four-part pipeline: PyPotteryScan pulls images and handwritten labels from documentation sheets, PyPotteryInk auto-inks pencil drawings, PyPotteryTrace vectorizes the shapes, and PyPotteryLayout assembles the final publication layout. The team tested it on 50 hand-drawn sheets covering 240 pottery drawings from the Terramara di Montale site in Italy. Participants in a usability study reported a median perceived speedup of 40x, with individual estimates ranging from 17.5x to 120x. The project's code is open-source, which matters in a field that rarely gets dedicated software budgets.
Ceramic documentation is one of archaeology's biggest publication bottlenecks: drawing, inking, and laying out pottery sheets by hand can take longer than the excavation itself. A tool that compresses that workflow without replacing the archaeologist's judgment calls could unstick a lot of backlogged fieldwork, especially at under-resourced institutions. The paper frames this explicitly as augmentation, not automation, since the human still decides what counts as a meaningful sherd feature.
Perceived speedup is not measured speedup, and one site's dataset is not a field-wide benchmark. Still, even a fraction of 40x would be a big deal for a discipline still drawing pots by hand in 2026.