Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PyPotteryInk: One-Step Diffusion Model for Sketch to Publication-ready Archaeological Drawings

Lorenzo Cardarelli

TL;DR

This work tackles the laborious task of converting archaeological pencil sketches into publication-ready ink drawings. It introduces PyPotteryInk, a one-step diffusion-based image-to-image translation pipeline using a patch-based, single-pass architecture with LoRA fine-tuning and a fixed prompt to produce standardized inked drawings. The method is validated on Italian protohistoric pottery, achieving rapid outputs with strong preservation of morphological details and publication quality, as demonstrated by training convergence metrics and expert assessments. The approach is open-source, GPU-accelerated, and adaptable to new pottery styles, offering substantial time savings and broad potential for standardization across archaeological documentation workflows.

Abstract

Archaeological pottery documentation traditionally requires a time-consuming manual process of converting pencil sketches into publication-ready inked drawings. I present PyPotteryInk, an open-source automated pipeline that transforms archaeological pottery sketches into standardised publication-ready drawings using a one-step diffusion model. Built on a modified img2img-turbo architecture, the system processes drawings in a single forward pass while preserving crucial morphological details and maintaining archaeologic documentation standards and analytical value. The model employs an efficient patch-based approach with dynamic overlap, enabling high-resolution output regardless of input drawing size. I demonstrate the effectiveness of the approach on a dataset of Italian protohistoric pottery drawings, where it successfully captures both fine details like decorative patterns and structural elements like vessel profiles or handling elements. Expert evaluation confirms that the generated drawings meet publication standards while significantly reducing processing time from hours to seconds per drawing. The model can be fine-tuned to adapt to different archaeological contexts with minimal training data, making it versatile across various pottery documentation styles. The pre-trained models, the Python library and comprehensive documentation are provided to facilitate adoption within the archaeological research community.

PyPotteryInk: One-Step Diffusion Model for Sketch to Publication-ready Archaeological Drawings

TL;DR

This work tackles the laborious task of converting archaeological pencil sketches into publication-ready ink drawings. It introduces PyPotteryInk, a one-step diffusion-based image-to-image translation pipeline using a patch-based, single-pass architecture with LoRA fine-tuning and a fixed prompt to produce standardized inked drawings. The method is validated on Italian protohistoric pottery, achieving rapid outputs with strong preservation of morphological details and publication quality, as demonstrated by training convergence metrics and expert assessments. The approach is open-source, GPU-accelerated, and adaptable to new pottery styles, offering substantial time savings and broad potential for standardization across archaeological documentation workflows.

Abstract

Archaeological pottery documentation traditionally requires a time-consuming manual process of converting pencil sketches into publication-ready inked drawings. I present PyPotteryInk, an open-source automated pipeline that transforms archaeological pottery sketches into standardised publication-ready drawings using a one-step diffusion model. Built on a modified img2img-turbo architecture, the system processes drawings in a single forward pass while preserving crucial morphological details and maintaining archaeologic documentation standards and analytical value. The model employs an efficient patch-based approach with dynamic overlap, enabling high-resolution output regardless of input drawing size. I demonstrate the effectiveness of the approach on a dataset of Italian protohistoric pottery drawings, where it successfully captures both fine details like decorative patterns and structural elements like vessel profiles or handling elements. Expert evaluation confirms that the generated drawings meet publication standards while significantly reducing processing time from hours to seconds per drawing. The model can be fine-tuned to adapt to different archaeological contexts with minimal training data, making it versatile across various pottery documentation styles. The pre-trained models, the Python library and comprehensive documentation are provided to facilitate adoption within the archaeological research community.

Paper Structure

This paper contains 21 sections, 2 equations, 9 figures.

Table of Contents

  1. Introduction
  2. Research Aims
  3. Materials and Methods
  4. Archaeological drawings and digital inking
  5. Image-to-image translation and diffusion models
  6. Archaeological dataset used and general training process
  7. Fine-tuning and inference
  8. Code Repository and the PyPotteryInk package
  9. Results
  10. Training results
  11. The case study: Montale pottery
  12. Expert validation
  13. Single-Blind Discrimination Test
  14. Qualitative evaluation
  15. Hardware and software requirements, scalability
  16. ...and 6 more sections

Figures (9)

  • Figure 1: Training dataset for the '10k' model. Images are "compressed" into a squared box for training purpose
  • Figure 2: Training dataset for the '6h-MCG' model. Extensive data augmentation is used
  • Figure 3: Inference patching for an example image.
  • Figure 4: Training metrics for the 10k model.
  • Figure 5: Validation metrics for the 10k model
  • ...and 4 more figures