Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

The current model does not support capital Q and Z #44

Open
ohmycaptainnemo opened this issue Nov 27, 2020 · 6 comments
Open

The current model does not support capital Q and Z #44

ohmycaptainnemo opened this issue Nov 27, 2020 · 6 comments

Comments

@ohmycaptainnemo
Copy link

Hi,

When I give words with capital Z and Q in them, I get a complaint that the model does not have those letters as its trained letters. However, I noticed that your web demo has not issue with those letter.
Could you please share your model that includes those characters?

@JBisc
Copy link

JBisc commented Mar 9, 2021

I checked the trainings data, and there is no capital Z and Q. Is there a Specific solution for that problem?

@ohmycaptainnemo
Copy link
Author

ohmycaptainnemo commented Mar 10, 2021

Yeah I faced that issue too.
The model used for the online demo seems to have Z and Q. I ended up using the online model.
I believe the online demo stores the model on your pc (client side). So if you look at the code for the demo and find out what the model is called, you might be able to find the file on your pc.
The second option, which is not elegant but works, is to use Selenium and python and save the generated results by the model. This works if you are only after data generation.

@GirkovArpa
Copy link

The online demo has been ported to rust, and a demo is available here.

@bryandam
Copy link

Sorry for necro-posting but I've been playing around with this and came up against the same issue.
The currently online demo doesn't seem to have Q and Z and instead interprets Q as O and Z as either t or j depending on the style.

@ohmycaptainnemo: When you say training data, do you mean the data this project extracts from the IAM On-Line Handwriting Database or that the database itself lacks training for those characters?

@chavenor: I know you mentioned getting a Beam2 and trying to create your own dataset. Wondering how that went.

@ohmycaptainnemo
Copy link
Author

@ohmycaptainnemo: When you say training data, do you mean the data this project extracts from the IAM On-Line Handwriting Database or that the database itself lacks training for those characters?

It has been a while and I do not exactly recall how I specifically came to the conclusion that the training data was lacking. Perhaps there was a particular error message or notification that let me to that conclusion. I am not sure (and prefer not to speculate) what the source of the issue is, but I think my observation was as per what I mentioned previously.

@bryandam
Copy link

Thanks @ohmycaptainnemo; upon re-read I see I missed two things:
A) You were initially asking about and outright error. The project was throwing an error whereas the online demo 'worked'. Worked poorly mind you, but it works:
image

B) That it was @JBisc who mentioned the training data. I downloaded the IAM database and yea, it seems to be missing these, characters. Which sorts of makes sense, they were writing real-world examples instead of pangrams and there's not a lot of sentences that are going to start with Q, X, and Z.

I did find the model binary file from the online demo (b.bin) but I'm not exactly sure how I would use that in this project. Not that it matters really, as even if I could it wouldn't lead to output that's legible. Someone would need to extend the IAM database to make these work.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants