Image processing automation?

Last Sunday in my ward a senior couple reported about their service mission experiences. One of their tasks was apparently using a computer to manually place lines in images of obituary records to separate one record from another. Is there a technical group somewhere working on automating information extraction from images of historical records? How about other kinds of image processing or machine learning applications? I did a search of all the forums and didn't find anything immediately, and none of the technical projects seem at all related to that kind of work.

That would sure be a fascinating project to volunteer for, and I think it could do a lot of good if that kind of work is currently being done manually.

Sean

seanlittle wrote:Last Sunday in my ward a senior couple reported about their service mission experiences. One of their tasks was apparently using a computer to manually place lines in images of obituary records to separate one record from another. Is there a technical group somewhere working on automating information extraction from images of historical records? How about other kinds of image processing or machine learning applications? I did a search of all the forums and didn't find anything immediately, and none of the technical projects seem at all related to that kind of work.

Such work would fall under FamilySearch, which collects, organizes, and indexes genealogical images. There's a huge project underway for indexing obituaries, and this sounds like prep work being done so that obituary images can be indexed.

I've been thinking the SAME thing!

Especially now that many machine learning projects are open-sourced (Tensorflow, Torch, Theano, FAIR, etc), it would be amazing if someone with deep learning expertise could apply some of the image algorithms to indexing. Seriously, the indexing could be 99% automated with just human eyes for checking it off...

Then, the next application of machine learning is automated genealogy.

I only know the basics of machine learning, but would be SO happy to volunteer in any way possible on a project such as this.

Any progress on this? I am thinking about trying it here in a couple of weeks. This is in my realm as a robotics engineer and need to make some progress in using machine learning. What better place than with indexing...should be simple to start...

Related thread: Handwriting recognition software for indexing?