Polar Rose Blog – A blog about sharing photos & other stuff that matters to us
Our Private Beta Launch
As several people were quick to notice in the comments of older posts on this blog, the Polar Rose site changed on June 28th, the date we rolled out the private beta. While it was obviously a happy day for us here at Polar Rose, we’ve intentionally held back on posting any news since only a limited number of beta users have been let in.
Things have gone really well and while we will continue to be in private beta for some time, the time is now ripe to explain what can be expected in the near and mid-term. Here is a look at a simplified version of our internal roadmap document. We are currently approaching stage 0.2.
We will continuously roll out new features and refine existing ones in both the Firefox plugin and on the Polar Rose site according to the diagram above. We will also steadily increase the number of users we let in. While the first invitations went off to a targeted list of friends and critics, we will as of Monday (16th July) start selecting addresses at random from the nearly 28,000 that have signed up for notification.
“So what’s available now?”
Let’s be clear about one thing: the current version (0.18) of the Firefox plugin is not entirely stable and early beta users will have to take this into consideration. The primary issue is that if too many tabs are open in Firefox the plugin simply stops working properly and a browser restart is required.
We are going to spare you the explicit technical details about this issue, but we’re working with the Dojo community to solve this and are receiving fantastic support from wizards like Alex Russell, Greg Wilkins, and David Davis (thanks guys!).
With that limitation aside, the plugin accurately performs detection of faces in public photos, placing a rose approximately where the pinhole of a jacket would be. Clicking on a rose brings down a modal window with an input field which allows a user to name the person, agree to an existing name or edit an existing name. If there are multiple names from multiple users for a specific found person we show a list of “also known as” names. Your plugin interactions are shown on the Polar Rose site in a results list.
“What are the imminent features?”
Imminent upcoming features that we will roll out are:
- The ability to mark people as favorites
- The ability to mark people as “this is me” (ie. an image of yourself)
- A “grid view” of results to facilitate quicker browsing
- Tags & Comments
“So what’s not included then?”
As those of you who have been following us for a while will notice, there is currently no matching functionality. For instance detecting Metallica drummer Lars Ulrich with the plugin won’t bring up more photos of Lars. Polar Rose is “people-powered”, and unless our users discover photos and contribute (some) data we can’t bring qualified matches. However, the rate at which users are generating data through casual browsing is amazing. In only 15 days, 32 users have discovered 189,739 photos which contain 58,818 people and have added names to 3,668 of these people.
We will start adding automatic matching in “clusters” of photos. The first place you’ll will see a form of this in action will be in Search Agents that will have the ability to let you watch your friends Flickr or 23 photo feeds and will alert you to pictures with you appearing in them.
As our algorithms continue to improve with the data contributed by users we will will reach a point where we enable the matching functionality in the plugin for any detected person where we can provide relevant data. We can’t say when this will happen, it all depends on how quickly detected data grows, but with the numbers above as a guide we anticipate it to be sooner rather than later.
“What’s in the mid-term?”
As shown in the diagram above there are some other features that appear on our road map:
- Privacy functions to prevent us indexing images that users do not want indexed
- An adult content filter to enable users to view adult content in results only if they wish to do so
- RSS & Blog badges to enable the publishing of results on other websites
- Crawling of popular sites to automatically detect faces in images and further increase the the amount of detected data
- A version of the plugin for Internet Explorer
- A royalty free Javascript API to enable the broader web community to interact with Polar Rose
In closing…
If you have signed up to be a beta tester please be patient as we can’t accommodate all 28,000 requests in the near term, once that invite comes we hope you will have fun exploring and using Polar Rose, please do send us your feedback!
That’s great news, I’m curious though – how is this privacy thing going to work? I mean, are we talking about something as simple as minding robots.txt or will a user be able to block images on a case by case basis? If the users can block images then I’d be very worried about a denial of service attack going on, there are people out there that think that this is the end of the world for some reason and I wouldn’t put it past them to sit there and block every image they can. I for one would prefer to see no “privacy function” built in, because even in freedom/privacy loving America he have the understanding that you are not entitled to anonymity or privacy while in the street. The internet is the new street, this is a very interesting development.
Sterling, I thought you understood, but it seems to me you did not.
On privacy: We are consulting with various parties including some respected privacy organizations to find suitable and realistic measures. At this stage we do not have any concrete description and we will use the beta period to learn and understand how we can and will approach this. There is no simple answer, but we remain committed to continuously address this issue.
[...] The other day I came across a new startup Polar Rose is entering its private beta testing phase. For those who are interested in trying their early beta version, they allowed a limited number of beta users to test their plug-in. Polar Rose, is expected to launch its browser plug-in version 0.1 for Firefox in the next few weeks. It is a public photo discoverer, which means you can detects people in the public photos and you have the options to name the person, do some tagging as well on the edit the name of the photos that you found by using their browser plug-in. Starting from 16th July, they will randomly select some email addresses, i.e. those who sign-up for the beta notification to test out their plug-in. As I read their blog entry, they having some plans to further upgrading their plug-in, such as release the adult content filter, bookmarking feature, pubic API release, and etc. [...]
[...] Ein Update on Polar Rose: die Entwickler der Gesichtssuch – und Erkennungsmaschine berichten ausführlich über die jüngsten Entwicklungen. Konkret: was die Suchmaschine leistet, wie sie funktioniert, wo die Schwachstellen und Limits sind. [...]
:: Sigh::
Still waiting for my promised from two weeks ago Beta invitation :/
Oops! Got it, my bad! ::Blush::
This is what happens when you compute at 3am!
Blah blah “Human _ Powered” :)) I was thinking of a new tech . If you use people to do a work instead of some AI tech. This is just simply nothing, no new tech !!! I was right about this “new tech”.
You are using humans to do your work, not program or script. Can somebody explain to me, what is so revolutionary in your new searching tech ?
Slavcho,
you NEED people to tell the computer what name belongs to a certain face. If i show you a random picture you couldn’t tell who it is unless someone showed you another picture with this person and told you who it was. Only then would you be able to name the person in the picture I show you because you recognize him. THAT is the new tech :) Finding random pictures on the internet and if the system recognizes the face it tells you what the person’s name is. It’s on a per face basis, not per picture :)
What are you doing about two persons with the same name?
And what about the changes of the face in a live-time?
Whity
I agree about that and have one question. Can your program tell the name of someone if I show a picture of that person ?
I suppose that you program can’t tell who is this because if it can do that, you will need a supercomputer to work for face recognition. Because there will be a thousands of requests.
Slavcho,
That last statement was just silly, you need to look into how facial recognition technology works before making such far sweeping comments. The image is only analyzed once, then a number similar to a checksum is associated with that face. The server stores that checksum and image. When you’re computer encounters a new image of the same person it can do two things, either analyze the photo using your cpu or let the server handle that. More than likely it will be handled by your cpu, which doesn’t require a “supercomputer”, I did it thousands of times a day from a laptop in Iraq – no supercomputer involved. Then your computer polls the server with that checksum you came up with, very little cpu usage for the server.
No I don’t work for the company, I just use google.
you don’t want the service to be used by for example the police,
but what if they take a high res.-shot of a demonstration and put it online?
I hope you will find a good privacy solution!
Verified the many-tabs-open bug in firefox :-(
Pretty much stops my beta-testing, as I always have 10-20+ tabs open.
Suggest an easy way to switch the extension on/off as everyone probably don’t need the face recognising going on all the tine
Will there be a plugin for safari (mac/windows) ?
Kim: You can turn it on/off by clicking the rose at your bottom right and tick “enable PR”. We are working hard on the many-tabs bug, stay tuned.
Cougarten: Good points. We are committed to finding a good privacy solution, also for your example. Regarding change in appearance over a lifetime our algorithms won’t be able to cope with that. Your faces would be grouped into different versions of you. Regarding two persons with the same name, those can hopefully be separated based on appearance or other tools like tags, person naming etc
peter: Safari is after IE, priority wise, which means that there won’t be a Safari extension in the near future. :-(
Signed up right at the beginning, still no Beta invite, signed up again today, still no Beta invite… Will it ever happen?!?!?!?!?
Trevor: As described above in this post we are selecting from the list at random. You’ll get your email soon and you only have to sign up once.
Cmon with the face detection and matching already. That is the one and only reason I am interested in Polar Rose.
“Signed up right at the beginning, still no Beta invite, signed up again today, still no Beta invite… Will it ever happen?!?!?!?!?”
I have the same problem from the beginning. Maybe PolarRose just don’t like me :) .
I think this program will be great if it really can find someone from photo because I have so many photos on my computer and I don’t know who they are. I think that if the program can make some pattern from unrecognized photo and search that pattern over the internet maybe it can find another photo of this person and then it can be stored in database.
I am waiting for my beta :) .
And yes I am using latest firefox 2.0.0.6
When I install the extension (YES I’M IN!) it corrupts my interface.
Lol, ok it did and now it works (install – restart – disable – restart – enable – restart – works)
Waiting for invite too..
Thanks for the invitation. I hope that PolarRose will start a revolution in photo search.
I hope that issues specific to non-western naming will be dealt with e.g. In Japan typical name order is surname first then given name. Should Japanese names be entered in that order or in Western order? Maybe not as exciting as questions about cutting-edge face recognition algorithms, but it needs to be addressed. As does localization and translation, so that e.g. Chinese people can enter names using their own language and character set, but then western users can view the same pictures and see those names transliterated into Roman characters, and Russians can see them in Cyrillic, etc.. Does that happen now, or is it planned?
Hey, I got a beta code, but can’t signup.
What gives?
[...] Jetzt ist es also da: das Plugin von Polarrose für Firefox-Browser mit dem Gesichter in Bildern erkannt, getaggt, kategorisiert und auffindbar gemacht werden. [...]
Still no invite. I have checked my Spam filters and everything. Guys can you please give me some idea of when I can expect an invitation?
When will you have the plug-in for the Opera browser?
@chinmay, i’ve emailed you separately. let us know what the problem is and we’re happy to help.
/n
@Herman Krieger: we’re going to have to give a rain-check for that even though we love our brethren troublemakers at opera in norway. right now the priority is solving the last firefox issues and from there moving onto internet explorer.
sorry! :(
/n
Martin: You raise some valid points here. The “name” field was intentionally designed without first name/surname as two separate fields to allow flexibility in naming (something that is also reflected in the use of AKA lists). People should name in the format that makes most sense to them and to people that find that picture useful. The answer to your second question is: no, we do not currently convert any characters between roman/cyrillic/chinese but we should definitely start thinking about if and how we want to do that. If you have any ideas, drop me an email janerik@polar…
[...] je to několik dnů, či snad už týdnů, když jsem obdržel svou pozvánku do beta verze Polar Rose. I když se zatím jedná jen o ranou verzi 0.2x, již teď je plugin dobře funkční. [...]
[...] highlights of the release: * Stability! As those of you who have used the plugin know – and as originally reported – the plugin has had a longstanding issue of often losing connection and stopping to show [...]
i am really interested. only problem is the trouble scanning the pictures to search the face in the crowd. if it lives upto expectations it will really be a hit. all the best.keep up the good work