Offline-to-Online Handwriting Trajectory Recovery
Recovering pen trajectories from static handwriting images to help low-resource recognition and cross-language transfer. In progress.
Status: in progress, with Prof. Faisal Shafait (NUST).
My ICDAR 2026 work showed that pen dynamics help recognition. However, most handwriting exists only as scanned images, with no pen data. This project asks whether we can recover plausible stroke trajectories (order, direction, and timing) from offline images, and use them to:
- augment low-resource recognizers with pseudo-online signals, and
- transfer across scripts and languages where online data does not exist.
The current focus is on defining the recovery methods and an evaluation protocol that measures what matters downstream, not just how closely the recovered path matches the true one.
Urdu ligatures vary widely across writers, which makes stroke-order recovery hard.