Collect images of most frequently appeared 100 character classes

[yustoris]

Overview

• Currently, we have collected images for 50 of target 100 character classes, so go ahead to obtain rest 50 character classes.
• Goal is to collect 200 images per class

Progress

• [ ] Find workers
• [ ] Attempt to ask the workers to collect images
  • On going
• [ ] Scale out collection process
  • Try out cloud sourcing services like Amazon Mechanical Tank?
  • Attempt to active learning method?

[grosniko]

Hey, can I help with the manual work of labelling?

[yustoris]

@grosniko Sure, we'll welcome to work on gathering and labelling cuneiform character regions on the documents

[grosniko]

Great, do you have some sort of on-boarding kit? Maybe I can get other people to help.

I've been dreaming of an OCR smartphone app with which you could scan directly over tablets displayed in museums (such as the Louvre or in London) and get a realtime translation, so I was looking for solutions and stumbled upon your research. I'm not an assyriologist nor am I a good coder, but I can help with the manual stuff at least to make this happen. ᐧ

On Wed, 29 May 2019 at 15:36, Kenji Yamauchi notifications@github.com wrote:

@grosniko https://github.com/grosniko Sure, we'll welcome to work on gathering and labelling cuneiform character regions on the documents

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/project-abgal/hccc/issues/2?email_source=notifications&email_token=AJBI4CJWHTTD4UV4G3KAFX3PX2BFPA5CNFSM4E6IIGY2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODWPLE6Q#issuecomment-496939642, or mute the thread https://github.com/notifications/unsubscribe-auth/AJBI4CJJYAEIVR462BZCFNTPX2BFPANCNFSM4E6IIGYQ .

-- NIKOLAI NIKOLAEV FR +33 (0)6 71 48 96 34 https://mail.google.com/mail/u/0/#m_4458521527034035932_UNIQUE_ID_SafeHtmlFilter_UNIQUE_ID_SafeHtmlFilter_ @: nikolai.nikolaev@essec.edu

[yustoris]

@grosniko Sorry to late reply.

do you have some sort of on-boarding kit?

Currently, there is no good toolkit unfortunately 😞. However, I'm wondering if we can use OSS annotation tools such as https://github.com/tzutalin/labelImg . Do you have any idea or better tools/platforms ?

I've been dreaming of an OCR smartphone app with which you could scan directly over tablets displayed in museums (such as the Louvre or in London) and get a realtime translation

Wonderful idea ! Your idea fascinates me so well 😍.

Anyway, I would like to tell you more detailed request after I decide ways how to annotate characters on the hand copy images. If you have any suggestions, fell free to let me know.