Hi Scott, I stated that comment about picking up the phone in landscape mode badly. I meant to say that most VoiceOver users who use OCR apps, and lift their iPhones above the page, find it easier to lift it and keep it flat and stable, if they hold their phone with their hands along the long dimension (i.e. "landscape" mode). I do pick up and hold my iPhone in portrait orientation, and I can lift it and keep it fairly flat now, even with just one hand, in portrait orientation, and get pretty good results with OCR apps. (This is the result of muscle memory kicking in after training). However, if I were to teach someone, in person, how to use an OCR app, I might well have them rotate everything so that the document and phone are horizontally oriented, and so they can pick up the iPhone more stably in landscape mode.
For the extreme example, think about picking up a ruler, and being told to hold it flat in the air. It is extremely difficult to hold the ends of the short dimension and keep the ruler from twisting. Or think about carrying a long, thin, package. You'll always hold the rectangle by the long ends if you're carrying a stack of boxes and trying to balance them. The instructions for TextDetective do tell you to hold the phone in landscape mode: <begin quote> How to scan and read text 1. Select the “Scan” tab. 2. Place the phone in landscape and the home button aligned with the right or left edge of the document. 3. Recommendation: when pointing the phone at text, try to keep the phone straight and parallel to the text object. 4. Lift the phone away from the document-about a forearm’s length. Don't tilt or rotate, keep the phone straight. 5. Hit the “Start Scan” button and wait for the “evaluating” prompt. 6. Phone will vibrate for found detections. 7. Once OCR is complete, a text field with all the detected text will be shown. VoiceOver will speak all the results. 8. To read the individual results one by one, tap the "individual detections" button. In VoiceOver mode, use three finger gesture swipe left-right to navigate use three finger gesture swipe left-right to navigate through results (single finger swipe in non VoiceOver mode). 9. Hit Done-Back button to rescan. 10. Note: Text may be at the top of the document. If results aren't satisfactory, try placing the phone nearer the top part of the page, while still aligning the home button with the left or right edge of the document. Then lift the phone and scan again. <end quote> This combination of holding the iPhone in landscape mode while the document page has to be in portrait mode almost guarantees that you won't be able to get the entire text read at once unless you're reading a section of a page where there is only text in the central third, or so of the page. Or this will work for reading letter envelopes. There's another thing about holding the phone in landscape mode -- you can use the edges of the portrait page to check your alignment, because they're not so far from the edge of the phone. I think the fact that the developer of TextDetective was able to get National Institutes of Health grant funding to support some of the work in the app (see the app description page), and also bring down the cost of the initial app so that more people could try it and get started, was also a deliberate decision. I'd be really curious to hear the impressions from people who train others (and who have mastered using OCR apps themselves), about what the most effective teaching methods to use are. Just my thoughts. YMMV. Cheers, Esther On Jul 21, 2012, at 12:21 PM, Scott Howell wrote: > Esther, > > I thought the comment about most Voiceover users picking up their phone in > landscape mode interesting. I actually have always used mine in portrait > mode. I wonder though if the page has to be oriented correctly, I assume (and > this probably is a pretty silly question) Text Detective already assumes the > camera is oriented in the lower right corner, thus being 90 degrees off from > what may be considered correct orientation? In other words could people have > less than stellar results if the phone were oriented in the portrait mode? I > don't know if the instructions tell you to put the phone in landscape mode, > but would assume it does. I find this pretty interesting actually. > I do agree the simplicity of Text Detective does make it attractive, but even > as a 1.0 app I would have thought maybe the dev would have accounted for some > features such as using the flash or auto flash and orientation of the > document. I will be watching this app closely to see what comes next. > Interestingly, this app was mentioned in a message that went out to the > Section 508 working group in the agency where I work. The Section 508 working > group is an interesting name since the objective of the group is to deal with > accessibility of all IT and other technologies used in the workplace, > although we tend to stray into other areas. :) > Thanks for sharing the information. > > Scott > > On Jul 21, 2012, at 5:09 PM, Esther <[email protected]> wrote: > >> Hi Anne and Others, >> >> No, Text Detective does not yet support languages other than English. What >> that means for languages like French, which use a Latin alphabet, is that >> you'll get many of the words, but characters where you have an "e" with an >> acute accent will show up as the number "6", for example. >> >> The plus side is that it is simple to use, and it gives people some better >> immediate results because it has them hold the phone in landscape mode, >> which is the way most VoiceOver users will pick up their iPhones, and is >> easier to keep documents in alignment. However, because you're holding the >> phone this way, and the document has to be in portrait mode, you don't get >> the full page -- only a section of it. There's no setting to turn flash on, >> again in the interests of keeping things simple. Unlike Prizmo, >> TextGrabber, and Perfect OCR, you need to orient the page correctly to get >> results. The program won't work if the document is upside down or rotated >> 90 degrees as it will for these other programs. >> >> The design philosophy, I believe, is to let people get some results right >> away. If you don't get results, you simply rotate the page and try again, >> and because it's relatively fast to navigate the controls, it's not hard to >> do this. >> >> In my estimation, the actual time to OCR is not really faster than with >> Prizmo. The complication that I believe most people find with Prizmo, apart >> from having to practice the alignment and holding things flat and in good >> light, is that there are multiple steps, even though most of this is just >> double-tapping the "Next" button in the top right corner. And if Prizmo >> were to implement a "quick process" mode where it assumed all the defaults >> and that you'd make no changes at all (since you need vision to crop, adjust >> images, etc.), I bet that would make things a lot easier for VoiceOver users. >> >> With Text Detective you still have to hold things flat, aligned, and in good >> light, but there are fewer steps. It is also more tolerant of slight >> alignment errors, just as Prizmo is more tolerant of such errors than most >> OCR programs (believe it or not). >> >> For people who are having difficulty pressing the "Start Scan" button in >> Text Detective while holding the phone flat, try putting your forefinger on >> the button while you hold the phone in landscape mode with your two hands, >> but keep your middle finger free. Then, when you have lifted your phone to >> the correct height, use a split tap with your middle finger to press the >> button without shaking the phone. >> >> I think Text Detective has good features, but won't give full page >> information in its current mode. By the way, Perfect OCR supports Russian, >> although I haven't tested that out yet -- just the French. And TextGrabber >> gives the full range of ABBYY's OCR language support, so should also support >> Chinese, Japanese, Korean, Arabic, etc. in addition to the above. >> >> HTH. Cheers, >> >> Esther >> >> On Jul 21, 2012, at 10:33 AM, Anne Robertson wrote: >> >>> Having looked at the description of this app, I see no reference to other >>> languages, and yet it is in the French appstore. Does it support any other >>> languages? >>> >>> Cheers, >>> >>> Anne >>> -- You received this message because you are subscribed to the "VIPhone" Google Group. To search the VIPhone public archive, visit http://www.mail-archive.com/[email protected]/. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/viphone?hl=en.
