Skip to content

Latest commit

 

History

History
94 lines (72 loc) · 3.52 KB

README.md

File metadata and controls

94 lines (72 loc) · 3.52 KB
Raw

How to run UNLV tests.

The scripts in this directory make it possible to duplicate the tests published in the Fourth Annual Test of OCR Accuracy. See http://www.expervision.com/wp-content/uploads/2012/12/1995.The_Fourth_Annual_Test_of_OCR_Accuracy.pdf but first you have to get the tools and data used by UNLV:

Step 1: to download the images go to

https://sourceforge.net/projects/isri-ocr-evaluation-tools-alt/files/ and get doe3.3B.tar.gz, bus.3B.tar.gz, mag.3B.tar.gz and news.3B.tar.gz spn.3B.tar.gz is incorrect in this repo, so get it from code.google

mkdir -p ~/isri-downloads
cd ~/isri-downloads
curl  -L https://sourceforge.net/projects/isri-ocr-evaluation-tools-alt/files/bus.3B.tar.gz > bus.3B.tar.gz
curl  -L https://sourceforge.net/projects/isri-ocr-evaluation-tools-alt/files/doe3.3B.tar.gz > doe3.3B.tar.gz
curl  -L https://sourceforge.net/projects/isri-ocr-evaluation-tools-alt/files/mag.3B.tar.gz > mag.3B.tar.gz
curl  -L https://sourceforge.net/projects/isri-ocr-evaluation-tools-alt/files/news.3B.tar.gz > news.3B.tar.gz
curl  -L https://storage.googleapis.com/google-code-archive-downloads/v2/code.google.com/isri-ocr-evaluation-tools/spn.3B.tar.gz > spn.3B.tar.gz

Step 2: extract the files.

It doesn't really matter where in your filesystem you put them, but they must go under a common root so you have directories doe3.3B, bus.3B, mag.3B and news.3B. in, for example, ~/ISRI-OCRtk.

mkdir -p ~/ISRI-OCRtk
cd ~/ISRI-OCRtk
tar xzvf ~/isri-downloads/bus.3B.tar.gz
tar xzvf ~/isri-downloads/doe3.3B.tar.gz
tar xzvf ~/isri-downloads/mag.3B.tar.gz
tar xzvf ~/isri-downloads/news.3B.tar.gz
tar xzvf ~/isri-downloads/spn.3B.tar.gz
mkdir -p stopwords
cd stopwords
wget -O spa.stopwords.txt https://raw.githubusercontent.com/stopwords-iso/stopwords-es/master/stopwords-es.txt

Edit ~/ISRI-OCRtk/stopwords/spa.stopwords.txt wordacc uses a space delimited stopwords file, not line delimited. s/\n/ /g

Edit ~/ISRI-OCRtk/spn.3B/pages Delete the line containing the following imagename as it crashes tesseract.

7733_005.3B 3

Step 3: Download the modified ISRI toolkit, make and install the tools :

These will be installed in /usr/local/bin.

git clone https://github.com/Shreeshrii/ocr-evaluation-tools.git
cd ~/ocr-evaluation-tools
sudo make install

Step 4: cd back to your main tesseract-ocr dir and Build tesseract.

Step 5: run unlvtests/runalltests.sh with the root ISRI data dir, testname, tessdata-dir:

unlvtests/runalltests.sh ~/ISRI-OCRtk 4_fast_eng ../tessdata_fast

and go to the gym, have lunch etc. It takes a while to run.

Step 6: There should be a RELEASE.summary file

unlvtests/reports/4-beta_fast.summary that contains the final summarized accuracy report and comparison with the 1995 results.

Step 7: run the test for Spanish.

unlvtests/runalltests_spa.sh ~/ISRI-OCRtk 4_fast_spa ../tessdata_fast

Notes from Nick White regarding wordacc

If you just want to remove all lines which have 100% recognition, you can add a 'awk' command like this:

ocrevalutf8 wordacc ground.txt ocr.txt | awk '$3 != 100 {print $0}'
results.txt

or if you've already got a results file you want to change, you can do this:

awk '$3 != 100 {print $0}' results.txt newresults.txt

If you only want the last sections where things are broken down by word, you can add a sed commend, like this:

ocrevalutf8 wordacc ground.txt ocr.txt | sed '/^ Count Missed %Right $/,$ !d' | awk '$3 != 100 {print $0}' results.txt