Product Evaluations and Guides
KNFB Reader for iOS: Does This App Live up to All the Hype?
For several years, many people who are blind or visually impaired have been toting two cell phones: their new Apple or Android smartphone…and their old Nokia, which they keep around for just one reason--to use the KNFB Reader software. The KNFB Reader allows users to scan documents, fliers, work handouts, restaurant menus, and other printed materials on the go and with often startling accuracy. Until now, this software was only available for Nokia's Symbian operating system.
In September, concurrent with the release of iOS 8 and the new iPhone 6 and 6 Plus, K-NFB Reading Technology Inc. released the much-anticipated iOS version of their flagship product: the KNFB Reader. It's available from the iPhone App Store for $99.99, and it is supported on iPhones 4s and up, and on the iPod touch 5. It is not currently available for the iPad, though the company does plan to support iPads that have a camera flash in a future update. An Android version is in the planning stages, but details and release schedule have not been announced.
Since the early days of accessible iPhones, users have been bombarding the company with requests for an iPhone version of the software. But it wasn't until late 2013, after the release of the iPhone 5s and 5c, and, more importantly, the release of iOS 7, that the company finally announced their plans to develop and release an iOS version of the KNFB Reader. "We couldn't go backwards and release a product that wasn't as good as before," says James Gashel, Vice President of Business Development at K-NFB Reading Technology Inc. "Until the 5s and 5c, the iPhone cameras simply did not offer the same quality as the Nokia phones," he says. "Their low-light imaging and flash were not as good, and even more to the point, until iOS 7 we were blocked from critical camera control functions we would need in order to achieve quality text recognition."
I've spent the past few weeks using and evaluating the new KNFB Reader on both my iPhone 6 and iPhone 5. I will report my results and recommendations below, but first, a bit of history.
The Kurzweil Breakthrough
In the mid 70s Ray Kurzweil was working with MIT on pattern recognition when a chance conversation with a fellow airline passenger who was blind inspired Kurzweil to work on what he was soon calling his "reading machine." This was no easy task Kurzweil had set for himself, as he had to do the following:
- Invent the flatbed scanner in order to get a page of text into a computer.
- Improve optical character recognition. (At the time, the US Post Office was using a single-font OCR in order to process mail more quickly. For his reading machine Kurzweil would need to develop OCR that would work with an infinite number of fonts and font sizes.)
- Develop a text-to-speech engine so that once the printed page was turned into computer text, the information could be easily conveyed to a user with a visual impairment.
"We purchased the first six prototype machines for $50,000 each," says Gashel, who was in charge of the Washington DC office of the National Federation of the Blind (NFB) at the time. The first commercially available reading machines cost $30,000 and were about the size of a washing machine. Over the next 20 years the size and cost of the technology continued to shrink, until by the year 2000 several software-only options were available for about $1,000. Users would pair the software with their own personal computer and one of many low-priced flatbed scanners, and read the output with a choice of screen readers and text-to-speech engines.
In the early 2000s, as digital cameras and handheld computing devices began to take off, Kurzweil envisioned a new opportunity. "We created the first KNFB Reader and ran it on a Windows CE PDA," Gashel recalls. "We were able to sell the combined software/hardware unit for about $3,500. Later, when Nokia introduced their model N82 with a 5-megapixel camera, we were able to offer the KNFB Reader Mobile at $1,695 for the software, plus about $300 for the phone itself and another $300 for the Talks or Mobile Speak screen reader. A year later we reduced the price of the Reader software to $995."
Today, with the availability of the KNFB Reader for iPhone, the software's price has dropped by nearly 90 percent. And that price is only 0.0033 the cost of Kurzweil's original commercial reading machine.
The KNFB Reader Interface
The KNFB Reader app interface is elegantly simple. On the homescreen there is a toolbar across the top and a toolbar across the bottom. The remainder of the screen is divided vertically in half. The right side contains the "Field of View" button, which we will discuss soon. The left side contains the "Take Picture" button.
Double tapping anywhere in this screen area instructs the reader to snap a photo and automatically begin the OCR process. The recognized text usually begins self-voicing after only a second or two. But there's a lot that goes on in that brief time.
The KNFB Reader is exceptionally forgiving when it comes to picture quality. It processes the image to make the text more readable by the ABBI Mobile OCR engine. It straightens out tilted or crooked text, and compensates for wrinkled pages and curving text, such as the text on a can label. Of course the Reader can't recognize what it can't see. For me, cans led to mixed results, though I could usually determine what was in the can, and even pick up some of the nutritional information and recipes.
The Reader can turn most print into spoken text with just a quick point and shoot. Double tap the "Take Picture" button, or, if you tend to jiggle the phone when you tap, perform a split tap. Do this by laying two fingers on the Take Picture area, then lift one of them, which will produce a double tap.
The KNFB Reader also features a pair of useful controls designed to help those with visual impairments snap higher-quality pictures: Field of View Report and Tilt Guidance.
Field of View Report
We mentioned earlier the "Field of View Report" button located to the right of the "Take Picture" button on the app's main screen. This control helps ensure that the entire page will be photographed. Position your phone so you think you have all of the page you are scanning in view. Then tap this control. The app will snap a quick, low resolution image, then describe your camera's placement. For example, the app may report, "Right bottom edges are visible; rotated 10 degrees counterclockwise." Or, ideally, "All four edges are visible; rotated zero degrees clockwise."
I found this feature particularly helpful when I first started using the Reader. Its near-real time feedback taught me where and how far to hold the camera to get a good shot. Moving forward, I found myself snapping fewer and fewer Field of View Reports, partly because my picture-taking abilities had improved, but mostly because the Reader does such an excellent job pre-processing my still less-than-perfect images.
This control is a toggle located directly above the Field of View Report screen area. Its purpose is to assist you in determining when you are holding your phone level, both front to back and side to side. The more off-center you are, the faster your iPhone will vibrate and buzz. Slow it down to no vibration and you are ready to snap the perfect down-facing shot.
I found Tilt Guidance even more helpful than the Field of View Report. I quickly learned I have a strong tendency to tilt my phone forward and slightly to the left. Again, practice makes perfect, and though I cannot yet report that I can position my iPhone perfectly from the get-go, I do find it takes me much less time to adjust my grip.
The Tilt Guidance feature only helps correct your iPhone's downward shooting position. I do wish it also offered a second set of guidance alerts that would help me shoot more accurately straight ahead. This feature is also absent in the VoiceOver camera access features. Knowing I have two faces in my viewfinder is helpful--it would be even more helpful if I knew I was holding my camera on the straight and level.
Saving and Exporting Files
After a single or multipage document has been recognized by the Reader, you are given the option to save the file either with a descriptive name or the default date and time stamp name. The app saves both the image and the recognized text on your iPhone. You can access the file through the File Explorer tab and reread the document, view the original image, or re-recognize the image with new column or language settings.
The File Explorer includes tabs to call up your list of Scanned Files, PDF Files, and Image Files. (More about these last two soon.) Swipe down to one of the file names and use the "Edit" button to rename or delete the file. You can also export as a plain text file, a formatted text HTML file, or a KNFB file, which is an archive file containing both the original image and a text file. I can see this last being extremely useful for saving store receipts, legal agreements, and other signed documents whose originals you wish to store electronically with a text copy you can read with your screen reader.
Make your export selection and you are offered all of the usual Share options, including e-mail, Twitter, Dropbox, and other iPhone apps that will read that file type. One improvement I would like to see on the initial File Explorer screen would be the addition of Rename and Export to the Actions rotor menu. This would make these options significantly quicker to access.
Reading PDF and Image Files
We all know the frustration of trying to open a PDF file on a computer running Acrobat Reader from Adobe and receiving the dreaed "Alert! Empty Document!" message. JAWS users can use the Quick Scan feature to recognize any text screen-by-screen, and the upcoming Version 16 will allow full PDF recognition. If you own K1000 or OpenBook you also have the option to do a send-to-OCR engine print job, but using the KNFB Reader, you now have a new option.
Email the PDF to yourself, or highlight it in Dropbox, OneDrive or another file sync service's app. Use the double-tap-and-hold gesture to access the share sheet, then the "Open in" option to send the file to KNFB Reader. You will now discover an "Imports Available" message alongside the File Explorer button. Import the file, double tap to open it, then select "Recognize imported PDF."
I recently downloaded an appliance manual that was an unreadable PDF. Even the JAWS Quick OCR option could not find the text. The company's customer service department sent me an unprotected version of the manual, so I had something to compare. When I e-mailed myself the inaccessible PDF and had KNFB Reader recognize the ten pages, the app finished each new page in less time than it took to speak "Page 1 done, page 2 done…" Reading the first several pages and comparing the text to the unprotected version, I could find absolutely no recognition errors.
You can do the same with image files, but currently the process is rather limited and complex. I was unable to recognize an image from my camera roll, or a JPG I received via e-mail. In neither case was KNFB Reader in the Share or the Open In list. Instead I had to save the file into Dropbox, then open the Dropbox app on my iPhone. I had to make the JPG file a favorite, then double tap first the Share icon, then the "Open In" button. At that point, I got the same import notice in Reader's file explorer menu, which placed the image in the Images tab, where I could then recognize it.
I occasionally receive e-mailed JPG files in lieu of faxes. Hopefully, with Apple's loosening of their restrictions for Notification screen widgets and sharing, turning them into text using the Reader will be made much simpler in future releases. Dropbox is free for a limited account, but users should be able to open and recognize a JPG or other image file directly from within an e-mail, as they can with a PDF file.
Here's a quick look at many other useful KNFB Reader settings and features:
- Flash. The home screen offers the option to set the camera flash to on, off, or auto. Additionally, the Settings menu allows you to have the flash always turned on when you are in picture taking mode. If you have some sight, this option may work better for you, but the company advises against having both the flashlight and the camera flash on, as this will create washed out images that are more difficult to process and recognize.
- Batch Mode Processing. If you have several pages to scan, this mode allows you to take multiple pictures and have them all recognized together into a single document. I avoided this feature until after I had improved my picture-taking skills, but I now find it useful for scanning multiple pages of a brochure or other multipage document.
- OCR Language. The app's ABBI Mobile OCR engine currently supports 11 languages, including most European tongues. I've been advised that a future release will also include the ability to have a recognized document translated from one language to another.
- Document Type. KNFB Reader will recognize your scan as a single or multicolumn page. Single column will "decolumnize" the page and arrange the recognized text from top to bottom, from one column to the next. Multicolumn will read each line from left to right, which is how you would want to read most invoices--so the item and price information are next to each other--and spreadsheets where the information is designed to be read one row at a time. Unfortunately, in either case, the Reader is unable to export the text as a formatted document with tables you can later read on your PC using your screen reder's table or headings hotkeys.
- Profiles. This image capture screen option lets you choose between manual and automatic picture taking. Automatic is supposed to do just that--snap a picture automatically when the Reader sees text to be processed. For now this option seems to be more of a placeholder than a working option, which is to say: in its current state, it only very rarely worked for me. I strongly suspect that once the app is optimized for iOS 8 and for the iPhone 6 and 6 Plus, this option will become much more accurate in identifying text and when to capture an image.
My Experience So Far
Over the past several years I have tried a number of iPhone text recognition apps, including TextGrabber and Prizmo. I know other users have had a good deal of success with these apps, but none of them worked at all for me. As for the KNFB Reader, I have spent considerable time with the app, from the initial release to the latest version 1.2--and to condense my experience so far into a single word, that word would have to be "phenomenal."
These days one of the few remaining text objects with absolutely no e-alternative is the daily mail, and after installing the Reader, I headed straight for my mailbox. I plowed through a stack of bills, junk mail, circulars, and newspapers in no time, tossing the trash and reading bills and full-page letters, even with strong crease marks in the page. The Reader even did a stellar job recognizing the text beneath most envelope address windows, and I could figure out which catalog company had sent me their offerings encased in wrinkled plastic wrap. On some of the more difficult pieces of mail, the recognition wasn't perfect, but the results were nearly always at least as complete and accurate as I would get using K1000 and a flatbed scanner.
The Reader will only read printed text, not handwriting. It also failed to read currency, even the text with serial numbers and such. Also, when I snapped a credit card I could not get the number, since it was the same color as the card's background.
I had good success in my pantry with boxes and cans. The app even read a good chunk of the information on a bag of egg noodles. If the product name was printed very large, or in a highly stylized font, Reader could not pick it up, but there was almost always enough information to help me find that can of beans and not worry I might be about to open a can f chili instead. Nonetheless, I do still plan to keep TapTapSee, oMoby, and CamFind on my iPhone's home screen.
Pages from my laser printer were almost without exception recognized without errors. CD jewel cases and DVD boxes also read well, though the disk themselves would not read at all.
During an NFB convention demonstration, Gashel snapped a photo of a PowerPoint slide from 20 feet away and the app recognized it completely. I am blessed with not having a job where I need to suffer through a lot of PowerPoint presentations, so instead I printed "The quick brown fox jumped over the lazy dog!" in 72-point type and taped it to a wall. I could recognize the text with 100-percent accuracy from 10 feet away, but only when I was successful in centering the page in my view finder and holding my camera vertically, without angling it too severely in either direction. Here is where vertical tilt guidance would be a real help.
KNFB Reader did an excellent job recognizing the menu at my favorite barbecue joint, but menu recognition is not nearly as much a must-have as it was four or five years ago. These days nearly every eatery I frequent posts their menus online.
The same can be said for books. Between Bookshare, Google Books, and Amazon Kindle, rare is the book I wish to read that I cannot find in accessible form. Nonetheless I did try scanning a few books. Paperbacks were problematic, as the gutters between pages made it difficult to properly scan the first few and last few pages. I found it easier to hold the book and camera sideways, then shoot the book one page at a time. Larger paperback books were a bit easier, but at best the scans were not nearly as accurate as the results I get with K1000 and a flatbed scanner.
I also tried recognizing paperback and hard cover books using the Reader with a StandScan Pro, a portable stand that helps you position documents and your smartphone for optimal OCR results. This process was considerably faster, since my iPhone remained properly positioned above the StandScan's lens cutout and I could use both hands for most of the bookpositioning, only having to hold it with one hand for the brief time it took to reach up and double tap the "Take Picture" button. Page gutters were still a problem, however, and most hard cover books did not fit completely inside the StandScan and had to be scanned one page to the next. The StandScan did do a stellar job helping me scan my mail, offering up excellent results almost as quickly as I could slip each new envelope inside the stand.
One hundred dollars is a lot to pay for an app these days, and many may be put off by the price. But price is relative. If tomorrow OpenBook or K1000 were available for $100 for one day only, it would be considered the bargain of the year and people would be racing to log on and buy. That said, I do wish there were a demo version of this app available. I think a demo would spur sales significantly, as once you try this app I suspect you will wonder how you ever managed without it.
Product: KNFB Reader for iOS
Price: $99.99 from the Apple App Store
Available from: KNFB Reading Technology, Inc., 877-547-1500, firstname.lastname@example.org
Comment on this article.
Previous Article | Next Article |
Table of Contents
Copyright © 2014 American Foundation for the Blind. All rights reserved. AccessWorld is a trademark of the American Foundation for the Blind.