Open souaissa opened 6 years ago
They are different approaches to the Word classification (using seq2seq model). Seq2seqX is my own extended version of this model. But right now the CTC model seems to have better performance. For training of this models you will need data in same format as in folder data/words2
thanks Breta,
how to create data IAM words + file .txt
how to improve the image segmentation in words
The script in the scritps
folder should pre-process words from IAM dataset. For taining of CTC model you don't need .txt
files. These files are used only in Seq2SeqX and Gap Classifiers.
Hi,
what is the role of WordClassifier-Seq2Seq and WordClassifier-Seq2SeqX and how train it