How would you like to walk up to any object at all — no matter how foreign or unusual — snap a picture of it, and have your phone tell you what that object is? Well, there’s now an app for iOS and Android that lets you do that, and it’s called CamFind.
When I first heard about CamFind, I have to admit that I was incredulous. Optical Character Recognition (OCR) is established and tested technology, available throughout countless apps on iTunes or Google Play, but object recognition technology is a whole different story.
Text patterns are repeatable and recognizable, but recognizing some random object from anywhere at all in the world requires a significant library of potential objects, and an algorithm the likes of which most programmers would balk at.
So, does it really work?
Testing Basic Object Recognition
The claim on the CamFind website goes something like this:
“Simply take a picture of any object and CamFind uses mobile visual search technology to tell you what it is.”
That’s a pretty simple claim to test. So, I started with something simple: an empty Pepsi bottle.
I intentionally turned the bottle so the word “Pepsi” was nowhere to be seen. Still, just based on the logo, apparently, CamFind was able to recognize the Pepsi brand even before it recognized that the object was a bottle.
Not only did CamFind get the “bottle” part right, it somehow was even able to discern that it was a bottle made of plastic, and not glass.
When the app discovers what the object is, it doesn’t just tell you what it is, it also conducts an Internet search for you.
Of course, the shape of a bottle is pretty distinct. So, clearly that was one of the easier tests. What about a non-distinct shape, like a circle?
Digging through my pencil drawer, I discovered an old penny way in the back. Placing this flat on the desk, I snapped a picture of the penny using CamFind. Immediately, CamFind recognized the words “ONE CENT” on the penny. It took a little while longer for the app to recognize the object itself though.
But it did. The result was “ONE CENT PENNY”.
Next, I ran another test using another tiny object without any words on it. This time, CamFind impressed by recognizing not only the object, but also the color, responding with: “GREEN PAPER CLIP”. Image color recognition is actually something we’ve covered before at MakeUseOf.
Getting more impressed, I decided to really put turn up the heat. Digging through some old disassembled PCs in the basement, I pulled out a random circuit board. Actually, this was an old 1Gb stick of RAM. Bring it on CamFind!
First, CamFind recognized the color (I wasn’t as impressed this time around), and then after about 30 seconds or so of searching, it also recognize the object. It didn’t say something general like “circuit board”. It literally responded with “GREEN COMPUTER RAM”. Not Bad!
Using CamFind on Unusual Objects
So much for routine objects, I decided to crank things up a notch by using objects that were even more random and obscure. This time around, I grabbed a broken F3 keyboard key off the same pile of computer rubbish in the basement and tested CamFind on this square black thinga-ma-bob.
And, in less than 30 or 40 seconds, CamFind recognized the object using four words: “BLACK F3 KEYBOARD KEY”.
Undaunted, I next dug deep into the recesses of my junk drawer and located one of the most obscure devices I could find: a USB to HDMI video converter made by Startech. Confident that I’d be able to stump CamFind this time around, I snapped a picture.
To my amazement, each word popped up one at a time:
BLACK (No surprise there)
STARTRECH.COM (OCR technology at work… not bad)
USB TO HDMI (Say whaaaaat???)
That’s right, with little more to go on than the brand name and two black ends of a cable with different shapes, CamFind was able to accurately identify this obscure converter, and then it conducted a web search for the converter — providing me with a way to conveniently order more if I wanted to. All this from snapping a single photo.
So, it’s clear from these tests that even though CamFind is able to impressively identify some pretty non-standard objects, it does make heavy use of OCR technology whenever possible. Words provide huge clues as to what the object might be. So, in my final round of tests, I decided to snap human body parts with no writing at all on them.
CamFind and Human Body Parts
In this round of testing, I took a photo of the first thing I could think of – my foot.
This shoe actually does have writing on it (Sketchers), but it’s on the other side of the shoe. So, I waited patiently as CamFind searched through it’s archives and ran through its interesting algorithms to uncover what this strange object might be.
I was wondering if it would just say “human foot”, or maybe something as impressively specific as “Sketchers Work Shoe”.
Instead, the accuracy was somewhere in between. CamFind returned the words “BLACK WORK BOOT”, and then conducted a web search for it. That was pretty close. Not terribly specific, but at least it was correct.
Next up, I snapped a shot of my index finger. No words, no clues – just a human finger, not even attached to a hand. How would CamFind perform?
Pretty simple. Without any other clues, CamFind was able to sift through its impressive database of objects and correctly identified the image as a human finger.
By this point, I was more than impressed by the abilities of this app. I envisioned visiting antique auctions and snapping images of unusual objects to learn what they are. This could be a wonderful resource to have on a smartphone that’s always with you.
However, never one to be bested by the likes of a smartphone app, I made one last attempt to stump CamFind. Since I could see that it relied heavily on OCR technology to identify objects, I decided to write meaningless text (a phone number) onto white lined paper to try and confuse the app.
Maybe it would see the pattern of numbers and report back “phone number”, or just “numbers”. Or would it actually pay attention to both the text and the object and say “piece of paper with a phone number”?
Well, the final result was once again somewhere in between. The app was smart enough to ignore the phone number, and simply report back what the object actually was.
Using CamFind in Your Life
You may wonder what good this kind of app might be in your own life. Well, if you think about it, there are a lot of really useful things you can do with it.
- Need to re-order food in your kitchen that’s running out? Snap a picture and let CamFind call up a web search where you can click on the Amazon link to re-order.
- At a yard sale or flea-market and not sure what an interesting object is? Snap a picture and let CamFind tell you.
- Do a web search on interesting exhibits at museums.
- Use it on the job to order replacement parts and equipment when you don’t have your computer with you.
- Use it at parties and challenge your friends to try and stump CamFind!
A nice feature of the app is that once you identify an object with CamFind, it stays in your library of searches. Not only are the results stored, but so are the photos you’ve captured, so you can refer back to the information later without having to bookmark anything.
CamFind is one of those apps that you may not realize you even need until you download and start using it. After a while, you may wonder how you ever got along without it.
What Do You Think?
Do you have any other interesting uses for an image recognition app like CamFind? Do you know of any other apps for iPhone or Android that do the same thing? Share your own insights and thoughts in the comments section below!