Scholarlink

Historical maps are valuable for history, social sciences, and linguistics but pose challenges for automatic transcription. This competition edition continues to address detection, recognition, and linking of text in historical maps, with new features: expanded French Land Registers data, a new Taiwanese dataset with Chinese characters, synthetic training data, and improved linking evaluation metrics. Seven teams participated with over 25 submissions across four tasks and three datasets. While detection performance is strong, recognition and linking remain difficult, though improvements were seen with Bézier curve line fitting and enhanced linking pipelines. All resources are publicly available on Zenodo ( https://zenodo.org/communities/icdar-maptext ).

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ICDAR 2025 Competition on Historical Map Text Detection, Recognition, and Linking

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yijun Lin,
  • Solenn Tual,
  • Zekun Li,
  • Leeje Jang,
  • Yao-Yi Chiang,
  • Jerod Weinman,
  • Joseph Chazalon,
  • Edwin Carlinet,
  • Julien Perret,
  • Nathalie Abadie,
  • Bertrand Duménieu,
  • Ta-Chien Chan,
  • Hsiung-Ming Liao,
  • Wen-Rong Su,
  • Mengjie Zou,
  • Tianhao Dai,
  • Rémi Petitpierre,
  • Beatrice Vaienti,
  • Frederic Kaplan,
  • Isabella di Lenardo,
  • Youngmin Baek,
  • Michael Hentschel,
  • Yu Nakagome,
  • Ichimura Shuta,
  • Jeongtae Lee,
  • Chankyu Choi

摘要

Historical maps are valuable for history, social sciences, and linguistics but pose challenges for automatic transcription. This competition edition continues to address detection, recognition, and linking of text in historical maps, with new features: expanded French Land Registers data, a new Taiwanese dataset with Chinese characters, synthetic training data, and improved linking evaluation metrics. Seven teams participated with over 25 submissions across four tasks and three datasets. While detection performance is strong, recognition and linking remain difficult, though improvements were seen with Bézier curve line fitting and enhanced linking pipelines. All resources are publicly available on Zenodo ( https://zenodo.org/communities/icdar-maptext ).