ScrollPrize/villa

The Vesuvius Challenge monorepo

What it solves

This project provides a suite of tools and libraries designed to read ancient Herculaneum scrolls using machine learning and computer vision. It addresses the challenge of virtually unwrapping carbonized papyrus and detecting ink on the surfaces of these scrolls from CT scans.

How it works

The project is organized into several specialized components:

  • Data Access & ML: The vesuvius library handles CT scan access and the training/running of ML models for surface prediction.
  • Ink Detection: Specialized training and inference tools are used to detect ink based on the winning model from the 2023 Grand Prize.
  • Unwrapping Pipelines: Three distinct approaches are used to extract papyrus sheets from CT scans:
    • VC3D: A semi-automatic segmentation pipeline.
    • lasagna: A PyTorch-based optimization framework for refining papyrus surface meshes and driving fiber tracing.
    • Spiral Fitting: A fully automatic pipeline that fits a global spiral to the scroll by matching traced patches and fiber skeletons.
  • Infrastructure: The foundation toolset manages datasets and cloud infrastructure, including photogrammetry masking.

Who it’s for

This software is intended for researchers, computer vision experts, and participants in the Vesuvius Challenge who are working to recover text from carbonized scrolls.

Highlights

  • Comprehensive Pipeline: Covers everything from raw CT scans to ink detection and final unwrapping.
  • Diverse Unwrapping Methods: Offers both semi-automatic (VC3D) and lasagna's mesh optimization and fully automatic spiral fitting.
  • Lagsna Framework: Uses PyTorch to jointly optimize stacked sheets for consistency.
  • Grand Prize Winning Tech: Includes tools based on the 2023 Grand Prize winning ink detection model.

Related

  • Dispatch
    Virtually Unwrapping the Herculaneum Scrolls: PHerc. 1667 Read for the First Time

    Researchers have used high-resolution X-ray microtomography and machine learning to virtually unwrap and read the entire carbonized scroll PHerc. 1667, revealing a Stoic philosophical treatise on ethics from the 2nd century BC.

  • Project
    PaddlePaddle/PaddleClas

    A comprehensive image recognition and classification toolkit providing ultra-lightweight models and industrial-grade backbones for high-performance vision applications.

  • Project
    frgfm/holocron

    A library of high-quality PyTorch implementations of recent deep learning tricks and model architectures for computer vision tasks.

  • Project
    deepdoctection/deepdoctection

    A Python library for document understanding that orchestrates layout analysis, OCR, and token classification to extract structured data from PDFs and scans.

  • Project
    jacobgil/pytorch-grad-cam

    A PyTorch library for explainable AI in computer vision that provides a wide range of pixel attribution methods to visualize and diagnose model predictions.