Author Topic: OCR is being tested - Please help  (Read 2350 times)

Randi

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 13168
    • View Profile
OCR is being tested - Please help
« on: April 13, 2016, 11:45:30 am »
Philip has asked for our help !

The team working on OCR is ready for some test data - and our experts are needed to provide that test data.

Typed images have been loaded for Farragut 1942 at https://www.oldweather.org/#/ships.


For more details see:
OCR testing
and
Farragut - Questions and Comments
« Last Edit: April 13, 2016, 01:24:40 pm by Randi »

Craig

  • Shipherd
  • Hero Member
  • *****
  • Posts: 3356
    • View Profile
Re: OCR is being tested - Please help
« Reply #1 on: April 13, 2016, 04:37:39 pm »
OK, you talked me into it.

I did one page but some of my row heights were too small. It didn't let me overlap and the drawing lines are too thick. I really don't like annotating.

The other annoyance is that, when transcribing, if you click on zoom-out because the magnification is too great it reverts to the original size when you go to the next box.

If someone else wants to annotate I can do some more transcribing.


Caro

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 7438
  • Our end is Life. Put out to sea. Louis MacNeice
    • View Profile
Re: OCR is being tested - Please help
« Reply #2 on: April 13, 2016, 04:48:43 pm »
Sorry Craig.
You only get pages to transcribe that you have annotated yourself.
It takes some getting used to, for sure, but it does get easier with practice.  ;)

Randi

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 13168
    • View Profile
Re: OCR is being tested - Please help
« Reply #3 on: April 13, 2016, 04:51:55 pm »
The trick to overlapping is placing the cursor in one of the blank columns.

To draw the following, I placed the cursor in column #10 - Reading At. Ther.


Craig

  • Shipherd
  • Hero Member
  • *****
  • Posts: 3356
    • View Profile
Re: OCR is being tested - Please help
« Reply #4 on: April 13, 2016, 05:00:09 pm »
Thanks for the tips, I'll try another one. Is there anything I can do to keep the resolution at the level I set it?

Randi

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 13168
    • View Profile
Re: OCR is being tested - Please help
« Reply #5 on: April 13, 2016, 05:01:45 pm »
Is there anything I can do to keep the resolution at the level I set it?

Unfortunately, no. It is a prominent item on the fix list ;)

Randi

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 13168
    • View Profile
Re: OCR is being tested - Please help
« Reply #6 on: April 16, 2016, 04:17:26 pm »
Somewhat off the subject of OW, but on the subject of humans and computers...

Gamers outperform computer at quantum task

Quote
...
"To me it is more than surprising -- it is really mind-blowing," says physicist Tommaso Calarco of Ulm University in Germany, who was not involved in the study.
...
The players outperformed the scientists' computer algorithm, coming up with strategies the computer missed. The researchers then fed the human strategies back into their algorithm to improve the results, thereby drawing closer to the quantum speed limit.
...

AvastMH

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 7409
    • View Profile
Re: OCR is being tested - Please help
« Reply #7 on: April 16, 2016, 07:32:27 pm »
Then there's a quantum of solace for the computers in that we are still here to guide them. They are so lucky to have us  ;) ;) ;D

Craig

  • Shipherd
  • Hero Member
  • *****
  • Posts: 3356
    • View Profile
Re: OCR is being tested - Please help
« Reply #8 on: April 16, 2016, 09:03:31 pm »
Gamers are entangled and in two states at the same time. No wonder they can beat a mere algorithm.   ;D

Danny252

  • Editor and Shipherd
  • Hero Member
  • *****
  • Posts: 775
    • View Profile
Re: OCR is being tested - Please help
« Reply #9 on: April 16, 2016, 10:31:00 pm »
You only get pages to transcribe that you have annotated yourself.

Somewhat off-topic, but wasn't the entire idea that by separating the two parts, separate people could annotate and transcribe? I don't see why they were left separated if that plan was then thrown out (unless the eventual plan is for people to annotate and the OCR to transcribe?).

Randi

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 13168
    • View Profile
Re: OCR is being tested - Please help
« Reply #10 on: April 16, 2016, 10:48:53 pm »
I have that question in a list of issues for the PTB from your earlier comment (http://forum.oldweather.org/index.php?topic=4495.msg121454#msg121454) ;)

I like to annotate and then transcribe, but I realize that not everyone will feel that way.

Craig

  • Shipherd
  • Hero Member
  • *****
  • Posts: 3356
    • View Profile
Re: OCR is being tested - Please help
« Reply #11 on: April 17, 2016, 12:45:03 am »
I agree, Randi. The way we do it in OW3 is good, especially with the help of Michael's automated box drawing script. It is disappointing that OW5 doesn't incorporate this feature. With regular spacing there should be no need to draw each row. If a row is empty this could be indicated simply by a toggle switch at the beginning of each row.

I am waiting for the "circle of death" problem to be fixed before I do any more OCR samples.

Randi

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 13168
    • View Profile
Re: OCR is being tested - Please help
« Reply #12 on: April 17, 2016, 08:24:21 pm »
Michael has an automated box drawing script for OW5 ;)
However, on Farragut reusing grids is working very well for me - although I still want to learn how Michael's script works.

Danny252

  • Editor and Shipherd
  • Hero Member
  • *****
  • Posts: 775
    • View Profile
Re: OCR is being tested - Please help
« Reply #13 on: April 18, 2016, 08:53:11 pm »
I have that question in a list of issues for the PTB from your earlier comment (http://forum.oldweather.org/index.php?topic=4495.msg121454#msg121454) ;)

I'm glad someone remembers what I've said, because I never do!

Randi

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 13168
    • View Profile
Re: OCR is being tested - Please help
« Reply #14 on: April 18, 2016, 09:02:04 pm »
 ;D