On Mon, Jun 7, 2010 at 10:45 AM, Rodolfo Alcazar Portillo <nospaze@xxxxxxxxx> wrote: > Well I had a somehow bad experience. Tried saving my firefox passwords, > but Firefox doesnt' have any stored passwords exporting form > (edit/configuration/security/stored passwords, if anyone knows how to > export them, please.), so I did an OCR to my firefox screenshots. > > The best I did was with gocr, due to its configuration options, making > tests of best configurartion options > > 1. used Dejavu fonts on my screen > 2. resized the image (bigger), letter height about 20 pixels > (uppercases) > 3. tried a lot of attempts playing with certainty and space sizes: > > for a in $(seq 95 100); do > for s in $(seq 12 25); do > gocr -s $s -a $a screen.jpg > echo s, a > read > # this read command allows you to explore your text on screen, > # read S and A and continue by pressing ENTER > done > done > > 4. did my final recognition for all my saved screenshots with the best > combination of A and S, cant remember them > > By the end, I had a 98% characters recognized, bad for passwords. > Neither tesseract nor Ocrad offered good results for me. In your case, > maybe tesseract would be helpful. Also unpaper. Thanks for that - although in my case the original text is an old document with poorly formed print from an old style typewriter (you know, the kind that had an inked ribbon, and embossed characters that flew onto the ribbon and paper when a key was hit!) I may try again but the results I got from the attempts with all the ocr applications I tried had a success rate around 5% which means that typing it back from scratch would be quicker than editing an OCR'ed document! Mike -- mike c -- users mailing list users@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe or change subscription options: https://admin.fedoraproject.org/mailman/listinfo/users Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
