historical-document-analysis

Implementation of PHOSNet and Pho(SC)Net for Word Recognition in Historical Documents. Implemented using Tensorflow 2.x

📜 [ICDAR 2021] "A Deep Deformable Network for Instance Segmentation of Dense and Uneven Layouts in Handwritten Manuscripts", S P Sharan, Sowmya Aitha, Amandeep Kumar, Abhishek Trivedi, Aaron Augustine, Ravi Kiran Sarvadevabhatla

Semantic Segmentation of Historical Documents using Deep Learning Architectures

SegClarity: An attribution-based XAI workflow for layer-wise interpretability in semantic segmentation

The largest publicly released line-level dataset of historical Arabic manuscripts — 14 books, 3,043 pages, 28,600 lines, with margin/insertion-anchor annotations for non-linear reading order.

An entire model pipeline of data ingestion, model training, evaluation, and segment management into a single CLI application for the Vesuvius Challenge.

Code and supplementary results for the paper "S. M. Unter, E. L. Hertel, DDD - A Diagnostic Dataset for Character Recognition and Detection on Ancient Egyptian Hieratic Characters and Words" at the ICDAR 2026 conference.