Opensourcing FamilySearch's OCR/HTR software+trained models?

Sky287

EDIT

I just found the Full Text Search Questions group which seems the most appropriate forum to ask this in, so I will post there. Unfortunately I can't seem to delete this post.

Hello!

I wrote this question in the general "ask a question forum" and did not get an answer. I am hoping this group might be more applicable or somebody will point me in the right direction.

Full question here:

https://community.familysearch.org/en/discussion/174341/opensourcing-familysearchs-ocr-htr-software-trained-models#latest

But basically - I have been working on a personal project that involves transcribing all the historic recorded docs for all the mining districts in Alaska, along with KARDEX information from the US Bureau of Mines and other data that contains mixed typed and handwritten data. I've spent the last year testing different technologies and training models. It's working decently, it's working better than any untrained software or software that doesn't charge per page transcribed. However, I recently came across some of the same data that had been transcribed by familysearch, via FamilySearch Labs, and I have to say that it is much better than mine.

I am wondering if there are any plans to open source the technology, or ways to contribute data and get that transcribed, or volunteer to help with the development of the technology, etc, that would allow us to use it to transcribe our own archival records.

Thanks!

sky

Robert Raymond

See the Full Text Search Feedback community discussion of this question at https://community.familysearch.org/en/discussion/174511/open-sourcing-familysearchs-ocr-htr-technology