Tuesday, September 22nd, 2009

Community based photo organization.

[Week 3] If lastFm has been able to understand the musical tastes of thousands of its users, and that delicious suggest you relevant websites based on your own bookmarks, there shouldn’t be any reason why it wouldn’t be possible to do the same sorting with pictures? The Ffffound website use their visitors tastes and attempt to reinvent photo browsing online. Ffffound is not the first one to tackle the image sorting problem of the internet.
Google image search tries to comes up with the most relevant images based on text content surrounding the image, but it always creates a huge amount of false-positive results. The image quality suggested by google is often very poor due to the fact that images are not evaluated individually by goole, only the main website’s notoriety (google calls it page rank) is taken into account.
Flickr’s adressed this organsational challenge differently in a very. interestingness concept is an elaborated algorithm to sort rate images rate rating with something adapted to each users taste. It’s based on your appreciation of images, not an average global appreciation; this makes a big difference in the way we sort data.
While many things get wider appreciation than others, the mainstream culture is rarely represent some of the favourite things of every individuals; they’re rather a mix of mildly appreciated things amongst all individuals. This is I think the reason why everyone complain about the mainstream culture, even if there are also partially responsible for it.
While it’s hard to avoid such patterns in static media like TV, Cinema or Newspaper, there’s no reason to stick solely to this global model in the digital world. The goole search results are still the same for all users, and many innovative companies still tries deliberately to organize the content in a top-down fashion. Digg lets user vote for (or against) content, and the more appreciated content gain in visibility. The ratio of positive votes agains negatives ones is more important in the beginning than the total quantity of positive votes. That’s already a big improvement compared to simpler rating system. The facts that top news are selected by user is also really interesting because an editorial bias has theoretically less chance to been observed on democratic news platform. The Digg Images website is especially interesting to be compared agains Ffffound because it put in perspective the real added value of Ffffound: the personalized content.
The content become more interesting to all users when it’s personalized. Another factor that shouldn’t be neglected when comparing the two approaches is the “troll effect” that happen when there’s one golden top-down structure. Since everyone would ike their content to be on top of the list, massive efforts are put into places to rise above others. The whole SEO (search engine optimization) industry is fuelling on these top-ranking promises. All these effort to highjack the normally functional system make it harder to discern the good from the bad. To counter the sneaky tactics of web marketing strategists, platform owner (google, digg, twitter, flickr, yahoo etc) elaborates new control and filtering technique, and it’s perpetuating the cycle.
Flickr’s own interestingness formula is mostly based on the photo popularity amongst all the community and therefor directly linked to global appreciation.
Flickr’s interestingness is in the spotlight on the flickr explore section. 500 users submitted photos are selected everyday to be part of this gallery. Since interestingness is one of the only criteria to help sort the interesting pictures from the millions of boring pet snapshots, many applications have been build on this concept using the flickr API. The FlickRiver site suggests to see a group or an album content ordered by interestingness and allow a smoother scroll through larger quantities of images by preloading extra images dynamically. For instance compare the view of the awesome bubble project photos using the flickr group(ordered by date posted), or using FlickRiver view (more interesting photos first, always according to flickr).
Here are two other example of the Flickr Api possibility by Mario Kligerman.

Monday, September 21st, 2009

High-Power handheld photo projector

[week 3] While browsing Arts Electronica website, I discovered a very ingenious project of Julius von Bismarck. He had the idea of projecting an image at distance using a telephoto lens and a 35mm slide inside a camera body. Instead of having a continuous projection, he preferred a bursting flash projection. The Fulgurator flash is trigered by other camera flashes firing around. Since both flash are synched together with the photocell, the projected image is clearly registered in the camera who triggered the fulgurator with its flash.

The Fulgurator blueprints

Just to better blend in the crowd, I who suggest anyone who plan on building such hi-power photo-projector to build it on the famous russian photo-sniper camera. You can’t go wrong with something designed for the KGB.


Monday, September 21st, 2009

Personas Project

Throught Rhizome, I just discovered this critical web project entitled Personas by Aaron Zinman.

This MIT effort explores data aggregation around individuals online. The site ask its users to type the name of someone. The then software then retrieve online content associated with this person, at least its name. Using language analysis technique, the program then attempt to summarize someone’s essence in simple categories by analysing text-content around this person. While more content is analysed, some themes and keywords repeat; and trends become visible.

Screen shot 2009-09-21 at 2.25.18 AM

The results are presented to the user in the form of a simple chart highlighting the main themes associated to the given name. While the results are obviously discussable, they more often than anything else make a surprisingly accurate portrait. This is especially true when the chosen keywords relate to only individual. If a searched name is too common, or used for different matters, the results of all these subjects blends into one thing making the main trends harder to identify amongst all the key categories. Also, if there’s too few search results for a given name, the analysis won’t be as effective since the sample is too small to show real re-ocuring trends.

While these graphics are a rather simple analysis aiming to summarize something very complex (someone’s personality), this example show in a really simple and communicative manner how it’s possible and easy to scan online content and instantly organize it. Such monitoring techniques are often used by authorities to better spot dissident behaviors. While the internet provide an ever growing sets of services and tools to better organize our digital life, it becomes extremely easy for content owners to analyze their users and therefore get a better picture of them. I found the Personas peice did a really good job at depicting how someone can be, accurately or not, labeled by automated systems.

Data mining at the scale we know it today is a relatively new trend; the way the internet has first been planned didn’t really imply such data-centralization. Originally, it was very anonymous, and really few links existed across the different aspects of one user. Emails, chat conversations, favourites site, browsing history, and personal photos were all things meant to travel over the internet but also ultimately remain located on the desktop computer. It’s still possible today for someone to remain fairly anonymous online by using the original means, but these usages slowly become deprecated as everyone is moving to toward a more convenient centralized network. In most cases, new ways of interacting online are by far more practical than the original ones.

Google, as everyone knows, is one of the biggest data mining company. But unlike many more discreet business, they did an amazing job at contributing back some very good analysis tool the community, and freely! Users can not only use desktop-like applications, but they can also explore the google database itself with various tools. For instance users can see which news is currently getting most media attention, or find out why certain keywords are more searched than others. By empowering their users in such ways, Goole let masses access and explore detailed data that would have overt-wise  been kept private, or only shared amongst an small elite. While it would be hard to argue that these shared bits of information can’t be beneficial to the society, I think this widely available data also contribute to make us forget about the remaining personal informations that lives in third-party databases. Without getting paranoid, we can easily imagine how detailed user profiles would be if we were combining and aggregating only few of these data source together:

  • Basic profile information (age, gender, location, political view, etc)
  • Friends network
  • Search and browsing history
  • GPS positioning and tracking (such system has been used for years in the transport industry to spot irresponsible drivers)
  • Professional information (LinkedIn CV for instance)
  • Items boughs online (ebay, music, books, movies, etc)
  • Events users have participated in (for instance, via Facebook event system + GPS positioning)
  • Public and private photo albums. Photos bear timestamps recording when it has been taken, some camera can even register the GPS coordinate, or street address the photos has been taken.
  • Who’s on user’s photos. Facebook let their users tag their friend in photos. Apple iPhoto even attempt to recognize your friend faces in pictures for you.
  • Emails and chat communications

Other links:


© Copyright 2018 by felix menard and their respective owners.
Running on wordpress, 960, and hosted at (mt)