jpuigcerver/Laia

Trained Model for testing on sample pages

mineshmathew opened this issue · 3 comments

Is there any trained model available so that I can run the HTR on a few English handwritten pages.

thanks

@jpuigcerver Thank you for sharing the model. I was able to run the laia-decode function on my line image. The output I get is as follows ( for a single line image)
1_1_0_line0_gray_128h 67 42 36 1 64 68 32 32 36 64 64 1 56 38 1 28 54 76 1 58 28 66 62 56 40 36 52 1 34 36 28 54 34

I guess what I get there as transcription is a sequence of label ids. Can you please tell me how I get the ascii/unicode characters from these ids. I suppose there is a symbol table., which you used to train the given model file.

thanks

I was able to get syms.txt file by running the data preparation steps for IAM ( egs/iam/ )..thank you. I will try with this list