Needs OCR data improvements

#14
by mrasker - opened

I’m extremely impressed with the structured OCR/document analysis of this model, but I’ve found some significant blind spots that I believe I can help with. I’ve sent an email to OpenBMB on a proposition for improving their training data in this department. If anyone from OpenBMB sees this, then please check your email; it should be pretty easy to find based on this particular subject.

OpenBMB org

Hi there! We are happy that you used our model, can you tell me about the e-mail address of OpenBMB you have sent
Let me have a check!

It’s the Gmail address that is attached to your (OpenBMB’s) GitHub profile. I have a GitHub project linked on the email; it gives a 404 at the moment because it’s a private repo. Contact me by email to discuss further.

Hi there! We are happy that you used our model, can you tell me about the e-mail address of OpenBMB you have sent
Let me have a check!

I’ve been reading more user reviews of the model, and their feedback reflects what I’ve sent in the email. Again, if you can, please take a look at it. I feel that much progress can be made with what I’ve made.

Cuiunbo changed discussion status to closed

Sign up or log in to comment