April 16, 2008

Chinese Photo Site Tuyuan May Do Facial Recognition. Or It May Just Put Boxes Around People’s Heads

Michael Arrington

33 comments »

I can’t really say much about Tuyuan yet since it’s in Chinese and there isn’t much information (translated page here), but it certainly looks like they’re trying to tackle the facial recognition problem that has destroyed many a startup.

We’ve seen Riya (now focused on ecommerce via Like.com), Ookles (never launched), and Polar Rose (in private beta for nearly a year), among others. Most recently Tagcow came on the scene, but it turns out it uses humans to tag photos, which tends to produce bad data.

Will Tuyuan be any different? We have no idea yet. But we’re contacting them to find out. More soon. Thanks for the tip, Orli.

  • Sphere It

March 29, 2008

Image Recognition Problem Finally Solved: Let’s Pay People To Tag Photos

Michael Arrington

78 comments »

Most people have thousands of digital photos sitting on their hard drive. And the vast majority of those photos aren’t tagged or searchable. Want to find the 300 pictures of your youngest son amongst 10,000 others? It’s not going to happen. Unless you’ve been diligently tagging and categorizing those photos over the years, and who does that?

The problem is obvious. The solution, not so much. A trail of failed startups have tried to tackle the problem with a fairly serious application of technology, including: Riya (now focused on ecommerce via Like.com), Ookles (never launched), and Polar Rose (in private beta for nearly a year), among others.

And now suddenly TagCow appears, which allows users to upload photos and have them tagged within a few minutes. The technology appears to be “magic,” meaning there’s no explanation of it.

If there’s a mountain in the photo, it’s tagged. A dog? yep. A yellow cup? Absolutely. It does people, too. Upload an image of a person and say who it is, and all other images you upload will be tagged with that person, too. The service also integrates with Flickr and will auto tag the photos you have on the service.

Thomas Hawk, the CEO of photo site Zooomr, tried the service and declared it “really, really cool,” although he wonders how it works.

The answer is, humans do it. I note that the TagCow site is careful not to say anything about the tagging process, and never use the word “automated” or anything else that would suggests computers are doing the work. Munjal Shah, the founder of Riya/Like, agreed, noting that it recognized a witch in Thomas’ photo - he says this just isn’t something a computer can do today.

I haven’t confirmed this yet. I’ve emailed the company for a description of how the service works but have yet to hear back. Until we do, I’m betting that humans are the taggers. Note that Google has effectively thrown in the towel and uses humans for this kind of work, too.

TagCow appears to be offering the service for free, so the cost side of the business may be a problem for them down the road. And the business is definitely a little sketchy. Worried about the privacy of your data? Just don’t click on their Privacy Policy or Terms of Use: “Privacy policy is TBD.” and “Legal stuff TBD.” Not exactly a way to build confidence.

  • Sphere It

February 17, 2008

Fred Wilson - Hypocritical, Wrong and Conflicted

Michael Arrington

122 comments »

Fred Wilson lit a fire today suggesting that certain bloggers need to step it up a notch to improve quality and be more like mainstream journalists.

A fair point if spoken generally, although I’d argue that the quality of reporting done by many bloggers today, at least in the tech space, is equal to or better than most mainstream journalism. I think this is particularly true when we’re talking about breaking, non-embargoed news, where contacts and inside sources matter more than having all the time in the world to think about, research, write and edit an article. His point, therefore, should have been that all news writers need to step it up a notch and aim for better quality, which is sort of like saying nothing at all.

Normally I wouldn’t take issue with the statement, except that it was partially aimed at us. Wilson specifically called out our Erick Schonfeld for his post on social gaming platforms, as well as Matt Marshall at VentureBeat for a post he wrote about Like.

Wilson’s first gripe is that Matt, in his post about Like, didn’t give enough credit to competitor ThisNext. His second - that Erick, in his post on Zynga and SGN, suggested that the “two companies are neck and neck like Hillary and Obama,” when “Zynga is almost an order of magnitude bigger.”

Wilson fully discloses his conflicts of interest in the post - that he is a friend to the founder of ThisNext and an investor in Zynga. At that point, of course, a lot of the credibility behind his opinions comes into question. The two bloggers he is attacking have no conflicts with these startups.

He fails to realize that both Matt (San Jose Mercury News) and Erick (Fortune, Business 2.0) are seasoned mainstream journalists who’ve made the crossover to blogging. So his whole argument about blogging v. mainstream media loses yet more steam.

In reading the articles, it seems to me that Matt did an excellent job of highlighting a recent surge by Like while still noting relevant competitors. Erick’s post, which I am more familiar with, is in my opinion above reproach. Erick notes the strengths and weaknesses of both platforms and suggests that developers will ultimately make a decision as to which, or both, they will join. Erick also interviewed Wilson for the post and quoted him in it.

So what this really comes down to is this. Wilson didn’t like the coverage. But instead of simply disagreeing with and rebutting the points made in the posts, he went after the reputation of the writers themselves. That would be inappropriate even if he was right. But the fact that he was both conflicted and wrong makes it inexcusable.

Wilson failed to uphold the very standards of integrity that he demands from others. He failed to contact Erick or Matt before writing, and didn’t seem to have the facts to back up his argument. In a twitter exchange between us on this issue, he defended his sloppiness on the fact that he’s a blogger, saying “if you are a blogger you can say what you think, once you become a journalist, you have a different standard.”

Now, frankly, I’m confused. Bloggers can say what they think, but journalists can’t? I think what he’s trying to say is that Erick and Matt are no longer bloggers and now need to hold themselves to a higher standard - one that Wilson explicitly doesn’t hold himself to. That sounds like hypocrisy 101 to me.

Also, in a comment to his original post, he says “Erick didn’t get it wrong…but i think he missed the opportunity to get it right.”

How can you be both wrong and right at the same time?

Wilson partially retracted his post in a follow up, saying that he was sorry for singling out Erick and Matt, and saying that he “didn’t mean to take a shot at either of them.” But he then goes on to say that the whole exercise was a good one, since it started this great conversation on the issue.

That’s no apology, Fred. An apology would include you admitting that both posts were well researched and well written pieces. And that it was wrong to attack the reputation of these writers just because the conclusions reached by them were different than your own.

One last note. In the comments Fred says it isn’t even debatable that SGN is not a real company. From what we hear on the street, some very high profile venture capitalists are willing to bet some serious money that he’s wrong.

Update: Mathew Ingram says I went a little too hard at Fred here. I don’t necessarily disagree. Fred tends to come at people pretty hard, so I went hard back. But some readers won’t know that, so it’s worth pointing out.

  • Sphere It

September 12, 2007

Hacks Make Their Way Into Yahoo Products

Michael Arrington

14 comments »

Yahoo Hack days are a lot of fun, and some pretty interesting stuff comes out of them. But a persistent question is whether or not they are much more than fun - and if any of these hacks ever make their way into actual products.

The answer, apparently, is yes. Tonight Yahoo is announcing two product feature launches that were originally created at Yahoo Hack Days. - Shop By Color and MapMixer.

MapMixer

MapMixer is a tool that lets users “pin up” their own image over Yahoo Maps. The two images are melded to create a hybrid version that can be saved and viewed privately or made public - users can also adjust opacity and perform other tweaks to make it look just right. The ideal use is to add a very detailed map to the existing, less detailed Yahoo map. The melded map can also be embedded in a non-Yahoo website. See images to right and below for examples.

Google Maps allows various types of annotations, but nothing exactly like this.

Shop By Color

Shop by Color is a new Yahoo Shopping feature that lets users search or narrow results by selecting one of 56 different color hues instead of typing the color in manually.

Like.com, which we’ve covered recently, also allows image searching with non-text as the input. What Yahoo is launching is a lot different, but it is exciting to see image search moving beyond purely descriptive text as the input. Images can be queried directly, whereas previously just the metadata around an image could be queried.

Both were developed at Yahoo!’s Q1 2007 internal hack day on March 23rd. Hayro Kolukisaoglu and Sundeep Tirumalareddy created Shop by Color, and Nimit Maru created MapMixer.

  • Sphere It

September 4, 2007

RockYou Integrates Like.com Image Search Into Slideshows

Michael Arrington

27 comments »

Last November, Munjal Shah made a fairly tough decision and did an about face on his startup, Riya. Instead of continuing to focus on Riya’s existing product - facial recognition and tagging of photos - the company took its core technology and launched an image search engine called Like.com.

Unlike other image search engines, Like.com uses photos as the query, returning similar images as the results. The company focused on ecommerce, particularly fashion items like handbags, watches, shoes, etc.

Fast forward to nearly a year later. The company is generating real revenue from sales on the site - Current gross merchandise sales are running at about $12 million per year (Like.com gets a small percentage of that as an affiliate fee in revenue). 1 million or so unique visitors come to the site each month.

This weekend photo widget startup RockYou started to integrate Like.com results into slide shows shown on the RockYou site (example). For now, results are limited to showing shirts on sale that are similar to the ones being worn by people in the photographs. Viewers can click through and purchase a shirt that look similar to the one their friend is wearing in the photos.

So far, so good. Shah says they are seeing an $0.80 CPM on slide show pages and sharing the revenue wtih RockYou. Other partnerships are ready to roll out.

Slide shows with Like.com results are only being shown on RockYou.com currently - due to issues with advertising on social networks (particularly MySpace), they are not included in the embeddable widgets. It’ll take a whole new round of negotiations before we start seeing them there, too.

  • Sphere It

November 8, 2006

Riya’s Like.com Is First True Visual Image Search

Michael Arrington

170 comments »

Silicon Valley startup Riya, currently a photo search company focusing on facial recognition, is making a significant strategic and product shift this morning. Riya will continue as is, but the company is leveraging the core technology to launch a new image search engine called Like.com (see our previous coverage of Riya here).

Like.com is image search. There are lots of other image search engines on the web today. But all of them only take queries as text, and compare those text queries to the meta data attached to an image file. This data is notoriously thin, and companies like Google are resorting to using human labor to attempt to add descriptive keywords to images stored on their servers. Even specialty image search engines like Pixsy have fairly thin meta data for images. And all of the existing search engines allow only text for search queries.

The Like.com engine takes both text and images as queries, something no one else does. To return results based on an image query, Like.com compares a “visual signature” for the query image to possible results. The visual signature is simply a mathematical representatioin of the image using 10,000 variables. If enough variables are identical, Like.com decides the images are similar.

What this means - If you see an image on the web, like a watch that Paris Hilton is wearing in the picture to the left, and use it as an image query, Like.com will return results showing watches that look very similar.

If you enter a text query, like “brown boots pointed toe,” Like.com will convert that query into variables in the visual signature and look for related image results. See screen shot below for the results from this query.

The site launching today returns results only for shoes, jewelry, hand bags and clothing. The service will expand over time to include other categories, but these initial categories represent a very large portion of consumer discretionary spending in the real world. With each result Like.com will also present a link to purchase the item, and their hope is to generate revenue from subsequent purchases.

A key feature that Like.com will be launching in the next month or so is an image uploader and a toolbar. Upload an image to Like.com to see similar results. Or, simply use the toolbar to use any image found on the web as a search query. Either way, Like.com will return results for similar items.

Robert Scoble at Podtech interviewed Riya CEO Munjal Shah on video in preparation for the launch. See the interview here, and a product demo here.

On a side note, Munjal has written a series of fifteen blog posts talking about his experience as a startup CEO. This is a very useful resource for new entrepreneurs. And given the length of this series, I wouldn’t be surprised to see Munjal publish this as a book at some point as well.


  • Sphere It