Description

This code repository is the official implementation of the paper "SmartExp: An Adaptive Data Expansion Strategy for Improving Handwritten Text Recognition".

Usage

HTR_Model

HTR baseline model files for training and testing.

Img2IMDB

Used to convert images in different formats (e.g. .PNG) into IMDB format for use in construction of datasets and subsequent training tests.

fid_test

For testing HTG methods.

Data

IAM

https://fki.tic.heia-fr.ch/databases/iam-handwriting-database

CVL

https://cvl.tuwien.ac.at/research/cvl-databases/an-off-line-database-for-writer-retrieval-writer-identification-and-word-spotting/

GW

http://www.iam.unibe.ch/fki/databases/iam-historical-document-database/washington-database

(Online: https://github.com/cwiep/gw-online-dataset)

HTG Methods

GANwriting

https://github.com/omni-us/research-GANwriting

WordStylist

https://github.com/koninik/WordStylist

VATr

https://github.com/aimagelab/VATr

Extra Evaluation metrics For HTG Methods

FID：

https://github.com/mseitzer/pytorch-fid

ToolBox

https://github.com/chibohe/text_recognition_toolbox

Acknowledgements

Thanks to the authors of the above resources for their contributions to the HTR and HTG fields.