Towards Writing Style Adaptation in Handwriting Recognition

TL;DR

This paper introduces a writer style adaptation method for handwriting recognition using a Writer Style Block and learned embeddings, improving accuracy in writer-dependent scenarios and enabling new writer embedding estimation.

Contribution

It proposes a novel Writer Style Block with adaptive normalization conditioned on learned writer embeddings for handwriting recognition.

Findings

WSB improves accuracy in writer-dependent recognition

Embeddings can be estimated for new writers

Domain adaptation with fine-tuning outperforms WSB in some cases

Abstract

One of the challenges of handwriting recognition is to transcribe a large number of vastly different writing styles. State-of-the-art approaches do not explicitly use information about the writer's style, which may be limiting overall accuracy due to various ambiguities. We explore models with writer-dependent parameters which take the writer's identity as an additional input. The proposed models can be trained on datasets with partitions likely written by a single author (e.g. single letter, diary, or chronicle). We propose a Writer Style Block (WSB), an adaptive instance normalization layer conditioned on learned embeddings of the partitions. We experimented with various placements and settings of WSB and contrastively pre-trained embeddings. We show that our approach outperforms a baseline with no WSB in a writer-dependent scenario and that it is possible to estimate embeddings for new writers. However, domain adaptation using simple fine-tuning in a writer-independent setting provides superior accuracy at a similar computational cost. The proposed approach should be further investigated in terms of training stability and embedding regularization to overcome such a baseline.