Today, we announce a brand new release of our API. The biggest news that it is much faster. Recognition times below 1s are now possible. We achieved this by optimizing the whole query path and stripping out unnecessary overheads. This also brings API users much more granular control over how much meta-data they want to store and retrieve with each recognitions result. Furthermore, we also offer an Android demo application, which makes development of a simple app much easier.

What’s new in detail:

• Amount of returned metadata can be chosen for each individual request. If you only need a bare image recognition, you don’t have to parse all those metadata anymore. And you get the recognition results faster as a bonus.
• Extensible interface. Lacking something in our API? Let us know, so we can rectify that. The API is ready to go beyond image matching and we have several things in the pipeline. Stay tuned!
• Quirk free. Our previous query API was built around our apps and some of our API users had to work around its special behaviors. No more hacks! Ever.

These new features allow for new kinds of applications, also especially in the field of augmented reality, thanks to the faster recognition.

For the Android demo app, it lets you set basic config such as group ids, compresssion etc. via a dedicated UI. This way you can test the parameters that work best easily, and then take over the code into your own project. The screenshots below illustrate some details of the app:

API demo-app: settings screen. Allows for API specifying request parameters via a graphical user interface

API demo app: result screen. Shows meta data for matched reference image, as well as paramaters such as timing etc,

Sample code, the android demo app, and the documentation of the new API functions can be found in our github account as usual:
https://github.com/kooaba/kooaba-api/tree/beta/query_api_v2

Here’s a quick video tour:

About the Lumia 800 itself, we think it’s great! It feels solid, and its beautifully-shaped case makes it a pleasure to look at. Using Gorilla Glass 800 x 480 ClearBlack screen, its deep black color (from the AMOLED technology – as seen in Samsung’s Galaxy series) gives it a really nice finishing touch.

So, download the Paperboy for Windows Phone 7 app here – and if you don’t have a Windows phone yet, go and get a Lumia 800 (and by the way: it comes in many colors, not only pink).

So, wouldn’t it be nice to have a thin wallet that slips easily in the front pocket of your jeans, or the pocket in your suit? I have to admit that I am a bit of a productivity geek, and recently I spent some time researching solutions to this problem. Here’s what I found: Basically, there are three options:

1. Optimize the contents of your wallet
2. Get a better wallet
3. Get rid of your wallet altogether

Optimizing the contents of your wallet

A normal wallet these days usually contains some cash (bills or coins), some plastic cards (credit cards, loyalty cards, …), and some paper snippets like receipts, photos, coupons, etc.

How can you optimize this? The most potential is probably with the cards. There are some loyalty cards you probably rarely use. But of course, if you don’t have them in your wallet all the time, you certainly won’t think of bringing them along when you need them. The straightforward solution is to take pictures of the cards that you don’t need very often and don’t need a magnetic strip to use. E.g. for most frequent flyer cards, you just need the number, and the gate agent can just type it in instead of swiping your card. The same holds true for certain loyalty cards. So what do you do to remember the number? Take a picture of the card with your phone, and archive it so that you can retrieve it easily when needed. This works really well with Déjà Vu!

Same story works for for receipts: snap a picture to save it, dispose of the paper version, and keep the (backed-up) digital copy by using Déjà Vu. To easily retrieve cards and receipts, just use tags, like the video below illustrates:

Another way to optimize i.e. limit the number of cards you carry around is to see if you can combine certain cards. Some frequent flyer cards offer combining it with a credit card, the same works for the Swiss half-fare card. Some progressive countries such as Estonia allow you to put credit for public transport on your ID (so you don’t need an additional card). In some places you can drive without your license, as officers can check your driver’s license with your id.

In fact, the country you live in can make a big difference in the number of cards you have to carry. In the US, your driver’s licence is your primary ID. In many countries, they are separate cards. The currency is another big factor: Do you have to carry a lot of heavy coins (Swiss 5 franc coins, arrgh), or do you have to carry huge stacks of paper money with little value (Indonesia, arrgh).

Get a better wallet

Some wallets are already really fat and heavy even when they’re empty. (Go and check at the wallet section at your local department store. It’s quite surprising). But there are some wallets that explicitly try to solve this problem. Of course the look of it should still be somewhat decent. Googling around, I found this wallet, which claims to be thinnest on the Earth:

The Slimmy wallet

I ordered one and have to say it’s quite nice indeed. The only disadvantage is that it doesn’t have a coin pocket. But if it had one, it would drastically increase its size, so you have to carry loose change in your pocket.

Another option I was recently pointed to is the AbrAsus wallet (hat-tip: Alex). (Funny enough it is also from Japan).

Here’s a pretty cool demo video for that wallet (in Japanese).

Actually, a really great (i.e. slim) wallet would make a good Christmas gift!

Getting rid of your wallet altogether

Of course, with the smartphone revolution (which added another thing to carry around in the first place) the hope is that the phone will replace the wallet altogether.

Mobile payment is on the rise, and the latest phones come with NFC, which allows not only for payment but also couponing, receiving digital receipts, etc. One such implementation is Google wallet.

The same NFC technology can be used for identification in public transport or even to serve as your national ID. There are still some privacy issues to be resolved for the latter, of course.

Alternatively, if it’s just about cashless payments on the go, you could opt for the cool kid on the block, Square.

Recent studies claim that will only take a couple of years until mobile (in-store) payment will take off.

Until we have reached a wallet-less society – do you have other tips to make your wallet slimmer? Let me know!

Update: just this Sunday the fat wallet problem was discussed in one of the larger Swiss Sunday Newspapers. By now you should be convinced that there is a problem. And solutions to it.

There’s the one with Merkel and Sarkozy. Another with Obama and Hu Jintao. One featuring Netanyahou and Abbas.

According to the Associated Press, Benetton withdrew an image of Pope Benedict XVI and Ahmed Mohamed el-Tayeb after the Vatican threatened legal action. The White House made it clear that it was against the idea of using the US President’s likeness for commercial purposes. Sure, Benetton has a history of provocative ads, but these take the cake.

We’re happy to be working with Benetton this weekend to make these ads even more interactive. If you’re a reader of the Austrian Kurier, you can take a photo of the Kurier to unlock a special coupon redeemable at Benetton Megastores in Vienna and Saltzburg from Friday, December 16th to Monday, December 19.

via: The Kurier’s article in German

ATMedia.at wrote about our collaboration

Our image recognition software is fun to play with even if you’re not in Vienna or Saltzburg this weekend. If you take a picture of any of the ads with Paperboy, (try it!) you’ll be redirected to the advertisement on your phone where you can watch the UNHATE Guerilla videos, enter a contest and see the rest of the ads via the UNHATE website. There, you can watch the film UNHATE, by French director Laurent Chanez, about the precarious balance and complex interweaving the drive to hate and the reasons to love. That’s a nice message for the holidays.

Try it:  Download Paperboy for iPhone or Android. Take a picture of a page in your favorite printed publication and you are taken directly to the corresponding digital copy on your mobile phone. Make sure though, the publication is supported because it doesn’t work on all publications (yet). Click here for the list.

But this is just part of our vision. We ultimately want to bridge the gap between print and digital completely and combine the best of both worlds. Connecting all those people with all that content brings an incredible frontier of new opportunities. Whether you’re reading an article, commenting on a blogpost, or researching a new topic, you are communicating with others. This is something that is very common online, or face-to-face, but simply isn’t possible directly through paper publications, until now.

Now, with your phone, you can comment on something you read offline, without using your favorite search engine to type in keywords to find a corresponding article. Just take a picture of the article with Paperboy. And it will link you to the digital version of the article. Comments and all. Think about that. Let it sink for a while…. It’s almost magic

With Paperboy, we feel that we are one step closer to having the perfect blend of print and online content. As you know, many of us still prefer reading articles on paper, but many of us also want to be able to do things online, such as commenting and reacting.

Leave a comment on printed articles

When taking a picture, you will see a new link at the bottom of the mobile page like this:

Click it and you can leave a comment

It will be shown directly in the list of the reactions of others!

Commenting on content is just one of the things that we will continue to improve as we bring readers closer together. We hope that readers will soon expect high quality digital extras to accompany printed content.

Paperboy supports an ever-growing list of of newspapers and magazines. (Here you will find a list of all of the publications.) We think it’s quite impressive!

If you’re a publisher, you should join today. It’s as easy as pie. All you need to do is upload a PDF to our system, when there is a new publication. (If your favorite paper or magazine doesn’t use Paperboy – send them a note and ask them to!)

Much more to come

2012 is going to be an exiting year. Commenting on articles is just one of the first elements of our plan to create an all-round reader experience. By the beginning of next year, for example, you can expect a major update on Paperboy that will bring mind-blowing improvements in both UI and UX. We will continue to bring the community together in ways that add value to how and what you enjoy reading, online and offline.

And just as important, we will continue to expand our coverage to include more print titles. Many more magazines have shown interest and will be making their publications interactive for their readers.

For those who can’t wait, in three steps

Step 1: Download the script

Head to https://github.com/kooaba/kooaba-api/tree/master/data_api/python and download both KWS.py and upload_client.py. Place them together into the same directory.

Step 2: Install Python

This step is optional if you already have Python 2.5, 2.6 or 2.7 installed. Head to http://www.python.org/getit/ and install one of the available versions. Version 2.7 is recommended, but 2.5 and 2.6 will do as well (you’ll need simplejson package when using Python 2.5).

Step 3: Upload your first image

Set environment variables KWS_ACCESS_KEY and KWS_SECRET_KEY to the values of your respective API keys and run:

python upload_client.py --create-item-in <GROUP_ID> --title "My image" MyImage.jpg

where <GROUP_ID> is number of your group.

That’s it! Now you just need to wait a while for our indexing system to pick up your image and you can start recognition queries of your image.

For those who want to dig deeper

Before we explore the script functionality, lets have a look at metadata model. You already encountered group, which is a partition in our system. One of them was assigned to your exclusive use and you’ll be uploading your data to it. Each group can contain arbitrary number of items. Item is a central object of the metadata model as all the metadata entries are associated with one of the items. Each item must have a title, but all other metadata are optional.

But item alone is not very useful if you can’t recognize it. That’s where attaching images to items comes to play. You can attach more than one image to your item. Whichever of them gets recognized, the same item metadata are returned. This comes handy for example when you want to recognize a product not only from a front side view, but also from a back side view (small tip for cylindrical packagings: upload product photos taken from multiple angles as you never know when the product faces the customer by its side).

Let’s say your group ID is 1 and you’d like to test with an image of Mona Lisa. The following will create an item with the image of the painting and a reference ID:

python upload_client.py --create-item-in 1 --title "Mona Lisa" --reference-id "paintings/6th century/da Vinci/Mona Lisa" Mona_Lisa.jpg

Created item 35552981
 added image 23396181
 activated image 23396181

The script prints the operations as they are performed. Especially the ID numbers are of interest. For example we can use the item ID to add some links related to our picture:

python upload_client.py --create-resource-for 35552981 --title "Musée du Louvre" --section "Links" --uri "http://www.louvre.fr"

python upload_client.py --create-resource-for 35552981 --title "Mona Lisa on Wikipedia" --section "Links" --uri "http://en.wikipedia.org/wiki/Mona_Lisa"

Is it time to recognize the Mona Lisa painting now? Lest check:

python upload_client.py --get-image-status 23396181

ACTIVATION_QUEUED

Not yet, we need to wait for a while… What about now?

python upload_client.py --get-image-status 23396181

ACTIVE

Great, time to run the query!

Eeeh… seems there is a typo in the reference ID, it states 6th century instead of 16th. Fortunately, editing metadata is easy:

python upload_client.py --update-item 35552981 --reference-id "paintings/16th century/da Vinci/Mona Lisa"

Updated basic item metadata

And when bored of Mona Lisa, it’s easy to delete the whole item (but be careful what you delete, once it’s deleted, it’s gone for good):

python upload_client.py --delete-item 35552981

Enjoy testing and if you have any requests or suggestions about our APIs let us know at support@kooaba.com.

To make it short: as an app designer you always must ask yourself: When do you invite your user to a shortcut and when to make a detour? And where’s the balance between the two?

Let me present you some shortcuts and detours of our apps:

While building Déjà Vu, we had the end of the camera-scroll-detour in mind. Do you know that problem? You took a picture of a wine label you liked  in a restaurant. Now, five months later, you decide to buy this wine for a special dinner. You go to your camera roll (yeah, you kind of forgot to put it in your album) and scroll and scroll and scroll… The Déjà Vu shortcut: Take the picture of the wine within the app. It gets recognized and categorized automatically. Five months later: Just go to your wine memos and buy the wine with the link provided.

___

Paperboy shortcut: You read a great article in the Washington Post you want to recommend to a friend. Without the app you make a detour: you tear out the page, put it on the pile of pendencies of your desk and search for it 3 weeks later when your friend comes to visit. With the app you simply take a picture of the page and tweet it, mail it or share it on Facebook. Your friend sees the high resolution pdf of your article instantly.

___

Déjà Vu detour: if you see a DVD cover of a great film you can take a picture of it with the app. First, there’s a shortcut: The film title gets recognized automatically and your memo gets put in the film category. Now comes the detour, as you can explore a load of automatically added links: read movie reviews, buy the movie online, get the google results without typing the name or watch the trailer and get lost in youtube (like my friend the designer).

Oh, and if you’re interested, here’s a little detour on the great wide Amazon (not the bookstore). When the land’s descending, there’s only one way to flow for the river: downward. When it’s flat even, the river doesn’t really know where to flow, so it starts to meander in loops and detours exploring the country. Our goal is to apply this way of flow for our kooaba products: get straight to the point AND explore the wide range of possibilities.

I just finished updating our maps view, which shows the location of the mobile queries in the backend analytics tool. Now it is much faster and no more limited to the last 300 requests. Also, it looks much better (see screenshot below).

Remember, the old implementation was straightforward. It showed a marker on the map for each query (i.e. for each time somebody tried to recognize something with the application). This was pretty easy to setup but soon, we noticed that the number of queries can be quite big (a cool problem to have). Moreover, the map would take very long to load. Also, our customers were asking for a better representation.

I always wanted to try out the google-maps-utility-library-v3, which is a great library offering lots of cool things that can be done in conjunction with Google Maps. From this library I chose the MarkerClustererPlus v3 in order to group queries that are close to each other, depending on the zoom level the user chooses.

Check out the result in the screenshot bellow:

Here, customers can see all time usage or just during a specified period of time on a map. The queries are grouped together in circles. The different colors represent the concentraion of queries around the indicated area.

From a technical perspective, the markers (JSON containing the latitudes and longitudes of the query) are loaded with Ajax and fed to the clustering algorithm, which will group them nicely on the client-side. If the user zooms in, the algorithm will recreate the clusters for the area he sees at that time. This happens very fast as it is all done on the user (client) side. Same holds for zooming out.

Using this library proved to be a much simpler and cleaner solution that initially expected, so I ecourage you to try it out for yourselves!

These other links are also worth checking out:

• Code examples on how to use the library in your project
• One simple and a more complex example
• A benchmark of the library’s clustering speed

It is how much added value is produced for the customer in a given period of time.

While designing the logo we wanted to keep the mystic impact of the déjàvu effect. So we decided to work with a vortex shape in an abstract way to make the logo work for print and all other media.

The image below shows a summery of some the steps taken towards the final product.

As often, simple is best don’t you agree?

Visit the website of Déjà Vu!