/IAM-Splitting

IAM Splitting

The IAM Handwriting Database contains forms of handwritten English text which can be used to train and test handwritten text recognizers and to perform writer identification and verification experiments.

During research of HTR task we faced some difficulties with comparing results using IAM database. After some investigations we untangled the thread and discovered that some authors of papers compared results using different distributions. So we would like to share found splittings of IAM database.

Split Train Valid Test
IAM-A 6161 966 2915
IAM-B 6482 976 2915
IAM-C 6161 940 1861
IAM-D 9652 1840 1861

Related Work

If you have found any paper/work with these splittings of IAM, we are glad to see issues with links on work for adding information in this section. Thank you :)

Name Splitting
Evaluating sequence-to-sequence models for handwritten text recognition. IAM-A
Handwriting Recognition with Large Multidimensional Long Short-Term Memory Recurrent Neural Networks IAM-B
OrigamiNet: Weakly-Supervised, Segmentation-Free, One-Step, Full Page Text Recognition by learning to unfold IAM-B
End-to-end Handwritten Paragraph Text Recognition Using a Vertical Attention Network IAM-B
Attention-based Fully Gated CNN-BGRU for Russian Handwritten Text IAM-D

Repo Authors & Contacts:

Thanks

Thanks a lot Théodore Bluche for publishing this resource with IAM-B splitting and Joan Puigcerver for publishing IAM-A splitting here. Also we would like to express our gratefulness for authors that opened this problem in paper.