Re: [tesseract-ocr] Newbie: wondering why a fairly crisp document has such low accuracy

2017-08-12 Thread ShreeDevi Kumar
With English you should probably get close to 99% accuracy. Is your png at 300 dpi? Which version of tesseract did you use? Which traineddata? ShreeDevi भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com On Sat, Aug 12, 2017 at

[tesseract-ocr] Newbie: wondering why a fairly crisp document has such low accuracy

2017-08-12 Thread Stephen Boesch
I printed out the "Welcome" page on my HP laserjet printer and scanned it in using .png . The quality is quite good. So I had been anticipating maybe 85%+ accuracy on the tesseract-OCR. I did not even bother to tally carefullly - but by eyeballing it seems about 50%.I had used all

[tesseract-ocr] Re: Accuracy decreases when a Region of Interest is used

2017-08-12 Thread Isaias Barroso
Hi. I think you can try some things like: 1 - Set Segmentation Mode to PSM_SINGLE_LINE. I don't know the wrapper but maybe a rectangle is got before apply OCR process. 2 - Get the image for you interest area and save it to verify if the coordinates are correct or if the isolated area are