Is training with IAM Dataset done?

[breta01]

Is training with IAM Dataset done?

[Breta01]

Hi,

I have not tested the training on the data produced by this script yet. The script should work if you provide a location of words.txt and words/ folder from in IAM dataset. It should produce a new folder from which you should load images for training models.

One known issue of the script is slow image normalization function, which I would like to update soon. Otherwise, it should work fine.

[yasersakkaf]

Thank you.
I will try it out and tell if any necessary changes are required or not.

[yasersakkaf]

The script worked fine. The normalized images are ready. Which classifier should I train using these Images.

[Breta01]

Try the WordClassifier-CTC.ipynb, just change the location in words loading function and it should work.

[yasersakkaf]

Thank you.
That would be for word classification. How do I train the Gap Classification model and Char Classification model using this Dataset

[Breta01]

If you have working word classification, you don't need gap + char classification. Gap + char classification is different approach to word recognition and you can't easily use IAM dataset for it.

Plus in theory the CTC word classification should be better.

[yasersakkaf]

Thank you
I got it,

[souaissa]

Hi,
could you please tell me the requirement modification to use IAM daset

[Breta01]

@souaissa check the last comment in issue #49. I mentioned there required changes. I don't have much time, it would be great if somebody could make the pull request with that.