search

githubharald / CTCWordBeamSearch

Connectionist Temporal Classification (CTC) decoder with dictionary and language model.
https://towardsdatascience.com/b051d28f3d2e
MIT License
557 stars 160 forks source link

Creating TFWordBeamSearch.so file for custom dataset #24

Closed sirison closed 5 years ago

sirison commented 5 years ago

@githubharald

I'm trying to embed CTCWordBeamSearch to SimpleHTR inorder to build TFWordBeamSearch.so file on my custom data we need to have matx.csv and gt.txt file , how do I generate these file for my data, can I use the same TFWordBeamSearch.so file generated on IAM data set that is in the repository , please advise.

githubharald commented 5 years ago

Hi, you simply have to compile (see README) the decoder and then copy the resulting ".so" file to the src directory of SimpleHTR (see README). Then, call the SimpleHTR script with --wordbeamsearch parameter. The mat and gt files are only needed for testing (i.e., you can ignore them).