<div align="center"> <img src="figures\obsd.png" alt="OBSD logo"/> </div> <h2 align="center"> <a href="https://arxiv.org/abs/2406.00684">OBSD: Deciphering Oracle Bone Language with Diffusion Models </a></h2> <h4 align="center">Haisu Guan, Huanxin Yang, Xinyu Wang, Shengwei Han, Yongge Liu, Lianwen Jin, Xiang Bai, Yuliang Liu* </h4> <p align="center"> <a href="https://arxiv.org/abs/2406.00684"><img src="https://img.shields.io/badge/Arxiv-2406.00684-b31b1b.svg?logo=arXiv" alt="arXiv" /></a> <a href="http://vlrlabmonkey.xyz:8225/"><img src="https://img.shields.io/badge/Demo-blue" alt="Demo" /></a> </p>

Overview

OBSD (Oracle Bone Script Decipher) explores diffusion models for decoding oracle bone inscriptions and translating them into modern Chinese characters. This repository contains the official implementation, datasets, and training scripts from the paper.

News

2024-08-14 🚀 OBSD has been selected as the ACL 2024 Best Paper.
2024-07-15 🚀 OBSD has been selected as the ACL 2024 Oral.
2024-05-16 🚀 OBSD is accepted by ACL 2024 Main.
2024-02-15 🚀 Source code released.

Resources

📄 Paper: arXiv 2406.00684
🎮 Demo: OBSD Online
🗂️ Dataset: Character Evolution Dataset
🖼️ Modern character images: data/modern_kanji.zip
🛠️ Data processing helper: data/process.py

Data Preparation

Arrange your dataset so that oracle bone images (OBS) and their corresponding modern Chinese characters share the same filenames.

Your_dataroot/
  train/  (training set)
    input/
      train_安_1.png  (OBS image)
      train_安_2.png
      train_北_1.png
      train_北_2.png
    target/
      train_安_1.png  (modern character)
      train_安_2.png
      train_北_1.png
      train_北_2.png
  test/  (test set)
    input/
      test_1.png  (OBS image)
      test_2.png
    target/
      test_1.png  (modern character)
      test_2.png

Configuration

Update configs.yaml so that all paths point to your local setup.

data:
  train_data_dir: "/Your_dataroot/train/"      # path to directory of train data
  test_data_dir: "/Your_dataroot/test/"        # path to directory of test data
  test_save_dir: "Your_project_path/OBS_Diffusion/result"
  val_save_dir: "Your_project_path/OBS_Diffusion/validation/"
  tensorboard: "Your_project_path/OBS_Diffusion/logs"

training:
  resume: "/Your_save_root/diffusion_model"    # path to pretrained model

Quick Start

git clone https://github.com/guanhaisu/OBSD.git
cd OBS_Diffusion

conda create -n OBSD python=3.10
conda activate OBSD
conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia
pip install -r requirements.txt

Training

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python -m torch.distributed.run --nproc_per_node=8 --nnodes=1 --node_rank=0 --master_addr=localhost --master_port=1234 train_diffusion.py

Monitoring

tensorboard --logdir ./logs

Evaluation & Refinement

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python -m torch.distributed.run --nproc_per_node=8 --nnodes=1 --node_rank=0 --master_addr=localhost --master_port=1234 eval_diffusion.py
CUDA_VISIBLE_DEVICES=0 python refine.py

Download the FontDiffuser weights from Google Drive before running refine.py.

Citation

@misc{guan2025decipheringoraclebonelanguage,
  title        = {Deciphering Oracle Bone Language with Diffusion Models},
  author       = {Haisu Guan and Huanxin Yang and Xinyu Wang and Shengwei Han and Yongge Liu and Lianwen Jin and Xiang Bai and Yuliang Liu},
  year         = {2025},
  eprint       = {2406.00684},
  archivePrefix= {arXiv},
  primaryClass = {cs.CV},
  url          = {https://arxiv.org/abs/2406.00684}
}

OBSD

Install / Use

README