Tze Chuen’s Weblog

Yesterday I received an email from Evernote saying that I was invited to beta test their app. Evernote is an app for both Windows and Mac platforms that allows you to keep both text and image notes in Evernote’s servers. You can then refer back to them when needed.

What attracted me about Evernote was not that it allows you to keep your love letters from your ex-lovers. It was the text in image recognition technology. Initially, I thought that this tech is built into their app. However, after using the app itself I was slightly disappointed because the text in image recognition is actually done server-side, meaning all those tech stuff is actually being done on Evernote’s servers and not right on their app.

How it works is every few minutes your app synchronizes with Evernote’s servers uploading changes (such as additions of new notes). It takes a while before those text in image recognition process is done on your newly uploaded pictures. I’m guessing every few minutes the servers undergoes the process of “tagging” all the newly uploaded images by users. So it’s actually a matter of luck how long it takes for your images to get synced.

However, I don’t think the “few minutes delay” thing matters because the use of this technology is to allow you to search images that contain words that you remember. Unless your memory span is less than a few minutes, then you shouldn’t be worry about this.

This is how Evernote’s Mac app look like.

To test out how Evernote recognizes text in images, I used the “New Snapshot Note” feature and took a quick webcam shot thru my built-in iSight. I took a picture holding up Uncle Toby’s “what I eat for breakfast when I’m running late for classes”.

Right here you can see that it recognizes the word “Apricot” very well and highlights the word Safari-style. Other words that worked was “Uncle”, “Yoghurt” and the first few numbers of best before date. Anything smaller than the word “Apricot” was not recognized. Fair enough, those words are too blurry anyway.

Next, I took a picture of my book shelf with my digital camera and made sure the picture wasn’t too blurry.

As you can see, the smaller words such as “Socks” could be recognized.

Here, you can see that even vertical words can be recognized! This holds true for most of the words that appear on this picture except for those that are half-shadowed.

Just for fun, I typed in all alphabets a to z in the search box to yield all alphabetical results. Here you can see what words were picked up by Evernote’s text in image recognition system. It’s not too bad for a consumer-based product!

Of course, this is just one of Evernote’s many cool features. If you’re interested to know more, head on to www.evernote.com!

P/S: Evernote is currently invitation-only beta.