- TinEye Multicolor Search on Hacker News
If you haven’t tried our color search, go ahead and try it now. Warning: this is highly addictive! We also release a color extraction tool whereby you can extract all the colors present in an image, give it a whirl in our lab as well. Color search 101 and color extraction 101 cover the basics of the color search and extraction you can use in the lab and the API has been released too, we call it MulticolorEngine.
- The story behind the TinEye robot
Many of our TinEye fans have asked us about the history of the TinEye robot, if the robot has a name, how was it created? and if TinEye itself was inspired by Brandon Sanderson’s Mistborn trilogy (it was not). So we thought to start off the new year we would reminisce with Stephen DesRoches about creating the TinEye robot. We invited Stephen to tell you the story from his designer perspective.
We love robots, machines, and all things mechanical. In fact, our TinEye office is filled with mechanical friends! A little history: back in 2000 when we were brainstorming cool names for our image recognition technology, it was no surprise that robots came to mind. Not just any robots, but those awesome tin toy robots of yore. Yes, futuristic robots trump our human abilities in so many ways: their tin arms are stronger, their tin brains are faster, and their ‘tin eyes’ are keener… Et voilà, the birth of TinEye! Well not really, as when we registered TinEye in the year 2000 – oh that has a nice ring to it – there was no TinEye, no reverse image search engine and no image recognition breakthroughs yet. There was a team, excited about changing the world of image search, and the rest as they say is history! So grab a coffee and let’s journey back in time with Stephen.
Talking about robots
It was 2003 when I first met Leila and Paul during a demo of their pretty amazing image recognition technology. The idea of searching for images with other images was a new concept for me but as this technology advanced and we (silverorange and Idée Inc) became friends, it was not long before we were working together.
Fast forward to when TinEye was being developed and prepared for the world stage, I was given the opportunity and challenge to create a mascot, a logo, and an identity for the service: reverse image searching. This brand needed to represent speed and efficiency. It also needed to be friendly and have a great personality. TinEye’s reverse image search was magical and I needed to create a brand to impart that magic. Given the team’s love for robots, magic and speed, we all very quickly converged on a robot. But what kind of robot?
Drawing a robot
TinEye is a fun brand but a mysterious one. While fun — it’s not childish. It’s fast, but comes with a level of mystery. Most of us could not explain how the TinEye technology works (well of course except the engineering team). We just accept and expect that it does. So how can we illustrate all of this with a single mascot brand?
Starting is always difficult. I can spend days simply thinking about possibilities before ever picking up a pen. As a first step, I allow ideas to come and go while randomly throwing everything down on paper. Here are some of those initial ideas.
And from that long list, we start to eliminate. Everything that doesn’t work for any reason at all has to go. One round at a time, we narrow the potential candidates down.
From round 1, some of the robots we had a hard time eliminating included these, specially the sorting robot at his desk:
Selecting just one
After much debate, the speedy one wheel robot won us all over. It was time to focus and expand on that single concept. How would this robot look in full color?
Adjustments and final revisions
The problem with the robot above was that he didn’t look fast. He was heavy looking missing the goals of speed and efficiency. Below is the finalized robot. With small adjustments to a slightly more football-player-like shape, the slimmer and lighter colours were necessary improvements.
Speed? Yes. He could out run all of us.
Efficiency? Yes. Flexible rubber-like arms great for picking images.
Magical? Yes. How do the floating parts stay together?
Friendly? Yes. Well, he sure doesn’t look evil.
Personality? Yes. Loads of it!
Playing with emotion
The eyes are very powerful. The entire mood of an illustration can be altered by changing only the eyes and nothing demonstrates this better than the Pixar film Wall-e. A story without words but full of emotion. This alone provides the opportunity to use the robot in unique ways throughout the TinEye site.
Dissecting the illustration
At the root of all illustrations, it’s simply a stack of individual shapes (mostly created with the pen tool). Here is a look at all those layers and a fun way to see how TinEye was created.
TinEye Services for Business
Now that we look back at 2012, We spent the better part of the year working on TinEye Services for Business. This gave me the chance once again to experiment with many more fun story lines… and colour.
And of course, TinEye Labs, a place to show off current and future technologies and projects.
If you know the TinEye team, you will know that they are super friendly to wildlife and in addition to robots have a passion for animals. This is of course awesome when your office includes a garden where wildlife can be observed, but perhaps not such a good thing when you are building a new product brand. In the midst of our robot designs we decided — as a team — to experiment with an owl.
In design – very much like software development – going from point A to point B is not always done through a clearly defined road. On the path to the TinEye robot, our strongest alternative contender was an Owl. To create the owl – there are many types of owls – we started by drawing a series of owls and eliminating the ones we did not like or were difficult to build into a character. A lot of work and efforts but this was necessary for consideration but proved and confirmed our robot was the right choice. Here is the mascot that could have been TinEye. We can’t imagine TinEye not being a robot!
- MulticolorEngine: a color search API!
It is finally here!
MulticolorEngine, the API powering our color search lab is now released. You have probably already played with our released color search lab and already experienced addictive color searching but if you haven’t, give it a whirl today. In our lab, you can search a 10 million image collection by colors. Not just one color!
And let us know what you think. Would love your comments.
And if you need a handy dandy little guide for color searching using the lab, we’ve got that too!
What’s exciting today is that our color lab is now completely powered by our MulticolorEngine API – which (drum roll) has now been fully productized and is available for licensing.
MulticolorEngine = Search by colors in API form
Our MulticolorEngine API can be integrated with any image search collection to allow user to search images by colors. Well suited for e-commerce and image centric website and requiring no technology infrastructure changes, this API’s features include:
- A color indexer that extracts and indexes all the colors in a collection of images. No manual tagging/keywording of colors is required.
- The ability to search using one or more colors, or to search for images matching the color palette in another image.
- A color palette generator which will find all the colors present in a single image, or a collection of images. Using this feature you could display all the colors you have available in sofas, chairs and tables, and then let a user filter their search to only display green chairs.
- Support for structured meta-data search. Allowing you to build interfaces that can, for example, find all products priced less then $50.00, in the ‘shoe’ category that most closely match a particular shade of yellow.
- Arbitrary meta-data searching. So if you have multiple collections, tags, and prices, these attributes can be searched for along with up to 5 colors.
- Support to provide a count of the number of products you have that match a particular color, allowing users to easily browse your collection by color. For example you could let users know that you have 32 different varieties of red, 16 yellow, and 66 black shoes for sale.
- Easy integration with your existing search technologies and development infrastructure.
- The ability to ignore solid or transparent backgrounds in images. This is critical for product images where the background of the image should not be considered a color during a product search.
Right now, you probably need to stop reading and head over to the lab to start color searching!
In addition to our MulticolorEngine API and the color search lab, we thought it would be fun to build a little lab around one of our API feature: the ability to extract colors from an image. This new lab takes your image and gives you a color list or color palette. It will display a color palette for all the colors identified in your image. Color extraction works for JPEGs, PNGs and GIFs. Colors are displayed in RGB or HEX values. Give it a try. We would love your feedback. This of course is just one of the many features included in the MulticolorEngine API.
And that’s not all: how about shopping for shoes by colors?oh well, for that you will need to wait until next week when we release a little Zappos lab that will allow you to search products by colors!
With this API release, we are excited to continue building our vision for an integrated image recognition platform. So please stay tuned for more API releases and news in the coming weeks. MulticolorEngine is brought to you by the tiny TinEye team. If you are interested in bringing more image search and recognition goodness to the world, join us. We are hiring.
MulticolorEngine was built with love and caffeine in Toronto (Canada).
And if you would like to find out more about MulticlorEngine here is all you need:
- The internet map illustrated by traffic and user activity.
Russian data-visualisation designer Ruslan Enikeev has mapped 350,000 websites and 2 million links from 196 countries according to levels of activity and the other sites visited by their users. Each website is represented by a circle. The size of the circle is determined by website traffic. The color of the circle is determined by countries (for example US is blue, Canada is purple). The gaps between the circles are determined by the frequency the users go from one site to the other.
From Ruslan Enikeev:
As one might have expected, the largest clusters are formed by national websites, i.e. sites belonging to one country. For the sake of convenience, all websites relative to a certain country carry the same color. For instance, the red zone at the top corresponds to Russian segment of the net, the yellow one on the left stands for the Chinese segment, the purple one on the right is Japanese, the large light-blue central one is the American segment, etc.
Importantly, clusters on the map are semantically charged, i.e. they join websites together according to their content. For example, a vast porno cluster can be seen between Brazil and Japan as well as a host of minor clusters uniting websites of the same field or similar purposes.
- Toronto GirlGeek: the algorithms edition!
I am really excited to be speaking at the next Toronto Girl Geek evening. I remember attending one of the first Girl Geek dinners in London in 2005. It was organized by Sarah Blow, the founder of Girl Geeks. That evening was pretty much magical: I met Robert Scobble and Maryam Scobble. I also met Hugh MacLeod, Ben Metcalfe, Henriette Weber Andersen and a lot of awesome attendees whose names I can no longer recall unfortunately! It was a surprising evening, full of technology discussions, blogging, changing the world conversations and great wine – I vaguely remember a wine sponsorship there! Next week the Toronto Girl Geek evening is all about Algorithms. And that’s of course something I am super excited about!
Not only am I speaking but TinEye will be hosting in our offices. I would suggest that you get a ticket, but I hear it is sold out! Can’t believe that there are that many people interested in hearing about algorithms!
Inspire more women and girls into a career in science, engineering or technology by supporting Girl Geek Dinners. Perhaps your company could host the next one?
- TinEye commercial accounts… the really-super-easy way
A while back we introduced a commercial version of TinEye: a paid search alternative for professional, commercial or high-volume users. While the free version of TinEye only allows you to do a limited number of searches per day and is for non commercial use only, the commercial version of TinEye allows you to purchase as many searches as you like–for commercial or non-commercial use.
We initially launched the commercial version of TinEye as an API only. Using it required integrating the TinEye API with your web service or application. However what our users may not know is that we also provide a user-friendly interface for commercial accounts.
This means that you can create a commercial TinEye account and use it to search for images in the same way that you are used to doing at tineye.com. Upload an image, or cut and paste a URL. There is even a separate browser plugin for commercial accounts so that you can right-click on any web image to search for it.
So let’s get started! Here is a step-by-step tutorial on how to get you or your company set up with a TinEye commercial account and start searching… the easy way.
Go to the TinEye commercial website. This is different than the regular TinEye website, and all of the commercial activities are completed there, including searching and checking your account.
Click the Sign up tab to sign up for a commercial account. Fill in all of your details, and if you are using TinEye commercially, don’t forget to provide the URL for your company website. You will receive an email verification as soon as you’re done.
Check your email for a message from TinEye and click the verification link! You will be asked to log in with the email and password that you just used to sign up.
Once you’re logged in, you’ll find yourself on the Welcome page. Don’t click away yet! There’s some good-to-know stuff here to help get you started. You can get back to this page from wherever you are on the site by going to About > Welcome (but you need to be logged in!).
Before you can start searching, you need to buy a search bundle (you need to be logged in to actually buy a search bundle, but you can see the pricing even if you’re not). Transactions are handled via PayPal or credit cards; as soon as you’re done you’ll be directed back to your account summary page on our site, which will show what you just purchased.
You’re all set! Click on the Search tab to get started. Then simply search the way you would normally do on tineye.com. Upload an image from your local drive to search for it, or point to a web image or web page by pasting the URL.
To make searching even easier, get the browser plugin for TinEye commercial accounts. It lets you right-click on any web image to search for it (currently available for the Firefox and Chrome browsers only). To install, go to the Search page and select the API plugin for your browser. Remember you can only see this page when you are logged into your TinEye Commercial API account!
Note: The browser plugins for the regular version of TinEye found at tineye.com/plugin will not work with your TinEye API account. You must install the commercial version to perform searches that will work with your prepaid search bundle.
And that’s it folks, happy searching!
- TinEye Index
Don’t want to distract you all by an angry baby, but that’s exactly what I am going to do after I tell you that we have added over 15 million images to the TinEye index bringing our index to: 2,155,840,279. Sit back, relax and click play. Enjoy! Want more, drop by!
- We love hackathons! Let’s go to HackTO! Join us.
We love hackathons at the TinEye HQ and this Saturday we will be participating in HackTO. HackTO is organized by HackDays. And HackDays = hack events across Canada. This weekend HackDays lands in Toronto with 10 API and a 100 attendees! Did you know that Toronto is the 4th best city in the world to get your startup off the ground? It is!
Why such love for hackathons? Because we can see that our entire tech and startup community benefits when awesome developers (and people) get together to hack, build and tinker. If you have not joined us at any of our hackathons, it is time to remedy that.
The idea behind HackTO is simple: a single day hackathon where developers build an application using a set of provided APIs and any other API they would like to use. Completed applications get a chance to compete for awesome prizes.
This time around, there are 10 APIs to select from including our MobileEngine and MulticolorEngine. You may already know the MulticolorEngine as Piximilar. We are launching a set of TinEye APIs in the coming weeks and of course rebranding all current APIs. So stay tuned for our next announcements and get a preview at HackTO. In the meantime:
This is the API that allows you to integrate multicolor searching for your own image database or extract colors from images in your database.
MulticolorEngine searches an image collection using one or more colors. We will be providing a two APIs for the MulticolorEngine:
- A read-only API pre-populated and loaded with 3 million Creative Commons images. You will not be able to add or delete images from this collection.
- A blank r/w API that you can load up with any images you like.
To get a feel for the capabilities of the MulticolorEngine, visit the Idée lab.
MulticolorEngine is a REST API with responses delivered in JSON or XML format. The main methods are:
- color_search: Used to perform a color searches against the collection and return any matches with scores. You can specify up to 5 colours with weights in RGB or hex format.
- extract_colors: Given an image, extract up to 10 colours and their dominance.
- extract_collection_colors: extract information about what colours exist within your whole collection, or a subset of it
- add: Used to populate your image collection.
- delete: Delete images from your collection.
MulticolorEngine also supports full meta-data searching but we’re recommending you not to get into this unless you really want to – it is still in beta!
MobileEngine uses mobile variants of the same algorithms that power TinEye. MobileEngine is designed to find a match for mobile phone picture in a collection of original/target images. It can find duplicate and altered versions of images using fingerprint based searching algorithms. It can find images that have been cropped, resized, rotated, flipped, obscured, blurred etc. Some examples of use are to match a book cover to a collection of covers. Same with CDs & DVDs, stamps, currency etc. It works with rigid flat objects but not 3D soft objects like clothing, fruit, or family pets!
The MobileEngine API can be integrated with any mobile application. For HackTO this API image collections is not pre-populated which means that if you want to develop and launch a mobile image search application integrated with this API, you will need to create a database of images for your search. The TinEye team and their documentation can assist you with that prior to the start of HackTO. Get in touch.
MobileEngine is a REST API with responses delivered in JSON or XML format. Each operation is completely atomic, meaning that given a request, other requests will wait for the first request to complete before they are processed. All responses from the API will have these keys:
- status: indicates the status of the request, can be one of ok, warn, fail
- method: the method that was called
- result: list of results for the method
- error: list of errors from the API
The main methods for MobileEngine are:
- add: Used to populate your collection. There is one shared collection for everyone to use, so please use a prefix for your image names if you want to make sure that they will not be overwritten.
- search: Used to search against your collection and returns matches. Optional search parameters are min score, max number of matches and horizontal flip.
- delete: Delete images from your collection by passing a filename.
Full documentation for the APIs will be provided at the start of HackTO just to keep things interesting :)
The full HackTO schedule is now online. And if the awesome TinEye APIs are not enough of a reason to get you to HackTO, how about the prizes:
- First prize: $2,000 (cash)
- Second prize: $1,500
- Third prize: $500
- TinEye Index Update
TinEye fans: we have added over 18 million images to the TinEye index. Happy searching. And don’t forget that if you manage a large image collection, you can create and submit a TinEye Image Map for TinEye to crawl your image collection!
- ArtScienceCamp in Toronto: join us!
Art Science Camp is an unconference organized co-presented by Hart House and Subtle Technologies. Every year Subtle Technologies in Toronto brings us the very international Subtle Festival. For 15 years the Festival has been bringing people together to promote wonder, incite creativity and spark innovation across disciplines. The Subtle symposium, performances, workshops, screenings, exhibitions and networking sessions provide a forum to explore ideas and pose questions at the intersection of art, science and technology. I am sure you will want to attend next year’s festival! But in the meantime, join us for Art Science Camp.
So what is Art Science Camp: It is an Art, Science and Technology unconference. A two day conference which will pique your curiosity and broaden your interest by intertwining art, science and technology in a series of peer presentations.
Art Science Camp starts on Friday February 3rd, 2012 at 7 PM and gathers artists, scientists, students, engineers, architects, designers and geeks. The Friday evening party is the venue for collaboratively creating a program of events to take place the next day. Everyone is encouraged to bring a crazy idea, a work in progress, or a vital topic for discussion, and to organize a session around it. Anyone interested a presentation or a discussion can claim a presentation spot on the schedule.
Last year, the first Art Science Camp included presentations by:
Eric Boyd – wearable electronics designer, Toronto Hacklab leader
Dan Falk – Knight Science Journalism fellow, popular science author
Like last year, Art Science Camp is going to aim at bringing together people who would not normally have conversations with each other, and create a space for surprising and serendipitous connections.
The first Art Science Camp last year sold out, and this year it is about to sell out, so if you are interested in getting a ticket, don’t delay. Registration is open and we have close to 100 attendees already.
ArtScienceCamp is looking for sponsors to make this year’s unconference unforgettable. If you can land a hand, help with sponsorships, sponsor lunch, dinner or drinks, please get in touch. An awesome sponsorship which helps defray the cost of this volunteer event starts at $250.
- Art Science Camp (#artscicamp)
- Date: Friday February 3rd at 7 PM to – Saturday Feb 4th at 6 PM
- Location: University of Toronto, Hart House, 7 Hart House Circle
- Registration: Open
- Cost: $10.00
- Available for sponsorship
[Photograph (c) Nasa Goddard Space Flight Center]