I have been a JAWS screen reader user for several years now. Out of all of the computer screen readers, I’ve always gravitated toward it because of its overall functionality and reliability.
A relatively new feature in JAWS caught my attention when it was first released a few years ago as part of JAWS 19. Picture Smart allows users to use an AI tool that scans images and provides text-based descriptions of them. As someone who works in marketing and communications, I occasionally handle images for various projects. Having the ability to quickly learn about what an image contains has been a big help for me in many ways.
One nice feature of Picture Smart is the ability to read text in an image. Of course, traditional photos might not contain much text, unless it appears on a sign, a shirt, or something along those lines. However, if you’re like me who works with graphics designed by others, it’s quite convenient to know what text appears in a graphic. I will say that like most AI tools, the text recognition is not 100% accurate. But on the whole, I’ve been pretty satisfied with this option.
Picture Smart also identifies the number of faces in an image. I often work with head shots and other images of individuals or families, so gleaning this info helps me ensure I know which photo is which if I’m trying to distinguish between a professional head shot of a woman, for instance, and a photo of a couple. Once again, there are factors like shadows, hats, or other things that can interfere with the interpretation of how many faces might be in a photo. It generally seems to be fairly accurate though.
Another cool aspect is Picture Smart’s ability to recognize logos or famous landmarks. For example, if someone is wearing a Tampa Bay Buccaneers shirt, it should be able to pick out the logo and acknowledge it. While I haven’t come across it much in my work, it can also tell you if a well-known landmark – such as Big Ben in London – appears in a photo.
Object recognition is another option in Picture Smart. I will say here that this seems to be one of the more challenging functionalities of all AI-powered image recognition software out there. Sometimes it’s spot on, and other times it is simply dead wrong. The good news is that you can at least get some context clues from an image if multiple objects are identified. Even if one or two is inaccurate, there’s a good chance at least some of the items will be identified correctly.
The bottom line is that Picture Smart is a very handy tool in many instances. I think it’s pretty comparable to the majority of other image recognition products out there. If you’re looking for the most accurate description of an image, I still think sighted assistance through Be My Eyes, AIRA, or the “old-fashioned” family member or friend remains the best option. Hopefully one day, these tools will become so powerful that you won’t always need a normal set of eyes to paint the best picture of what’s in front of you.