Sorry for the repeated post. I missed the message mentioning that posts need to be approved the first time , and after waiting around 12 hours assumed that my first post wasn't posted and wrote this post. I think that it will be better for everyone if replies are kept here.
Upon more testing, I noticed that the crop can at time affect the quality of the result. It is very rare, but I ran into a situation where "8" was recognised as "3". Once again, changing the box slightly could allow it to get it right again. It is really odd to me that despite the text itself not changing whatsoever, having more or less background area can make this much difference.. On Saturday, September 5, 2015 at 7:37:04 PM UTC+9, AxB wrote: > > Hello everyone. > > I just started using Tesseract-OCR 3.02 to recognise numbers only. > > The number themselves are *probably* in Futura Bold font, styled in a > particular manner (see images). > > Using the "digits" parameter, Tesseract-OCR would either get it perfectly > or fail completely (return a blank). > > After quite a bit of testing, it appears that it is the "crop" of the > image is what makes or break. For instance: > > > <https://lh3.googleusercontent.com/-I6vx1-5KxGY/VepwFvh_OmI/AAAAAAAAABw/kSXSI8qsJiU/s1600/Test1.png> > When poorly cropped as above, with quite a bit of horizontal and vertical > blank, the engine will always fail to return anything > > > > <https://lh3.googleusercontent.com/-8IMD05QoIYY/VepweKPrTxI/AAAAAAAAAB4/EFfQGgoD4CM/s1600/Test2.png> > A crop like this, with a some space for extra digits would fail in this > particular example, but succeed at time. > > > > <https://lh3.googleusercontent.com/--fH0jI8pEeQ/VepyLQAw6zI/AAAAAAAAACE/Qm22VlnbqGI/s1600/Test3.png> > > A crop like this, has so far always worked. > > > The problem is that I am capturing the image automatically and need to > cover for a range of at least 5-7 digits. > > I would never need to crop as badly as the first example, but I do need > more leeway than the last one allow. > > Is there anything I could try to make something like the middle crop work > better? > > Thanks. > -- You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/tesseract-ocr. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/4e989f13-de3f-4f84-8ef5-f48eadcdfa69%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
