Handwritten text generation and strikethrough characters augmentation

We introduce two data augmentation techniques, which, used with a Resnet-BiLSTM-CTC network, significantly reduce Word Error Rate and Character beyond best-reported results on handwriting text recognition tasks. apply novel that simulates strikethrough (HandWritten Blots) handwritten generation method based printed (StackMix), which proved to be very effective in StackMix uses weakly-supervised framework get character boundaries. Because these techniques are independent of the network used, they could also applied enhance performance other networks approaches recognition. Extensive experiments ten datasets show HandWritten Blots improve quality models.