Posted January 7, 2014 4:38 pm by with 0 comments

Tweet about this on TwitterShare on LinkedInShare on Google+Share on FacebookBuffer this page

red dress labelIf you search for a red dress on Pinterest, you’ll find plenty thanks to all of the pinners who were kind enough to properly label their image. Which means you’ll also find things like this mislabeled photo which is neither a red dress or a prom dress. I hope that’s not a prom dress. .

If you just search “dress” you’ll find more red dresses that weren’t labeled with the word red.

That’s why Pinterest just acquired Visual Graph, a company that specializes in visual search and image recognition.

I’m just looking for a red dress, but Visual Graph has a loftier goal in mind:

Our mission at VisualGraph is to connect these inspirational images together.   While a single image can captivate and inspire, a network of connected images allows for inspiration, exploration and discovery at scale.   Our approach is to combine the state-of-the-art machine vision tools, such as object recognition (e.g. shoes, faces), with large-scale distributed search and machine learning infrastructures.

And today we couldn’t be more excited to announce that we are joining Pinterest!

On Pinterest, millions of people are curating and sharing billions of Pins everyday.   And these Pins are more than just images — they link to contents that can inspire and enrich people’s lives.   We are excited for the opportunity to combine machine vision with human vision and curation, and to build a visual discovery experience that is both aesthetically appealing and immensely useful for people everywhere.

The sentence I bolded – that’s the tipping point for Pinterest. Right now Pinterest is all about curation. Users collect photos and categorize them based on their own needs. “Places I want to visit.” “Craft project inspiration” “My style.” But I’ve started to use Pinterest in a different way. I’ve started to use it in the same way I use the Google image search – to find specific things.

It’s a subtle difference on paper but huge in reality. I wanted to find an example of a mixed media canvas. I searched Google images then decided to try searching the same term on Pinterest. I was pulled in by Pinterest. I’m not exactly sure why. Maybe it’s the white space and added text? Maybe the images are better? Whatever the reason, for the first time I realized that Pinterest was a very viable image search engine.

There’s only one problem. The results come from the text not the image itself. So I know I’m missing out on a lot of great images that are poorly labeled.

skirt1Visual Graph will solve that problem. Here’s an example photo from their website. They asked their search engine to find all the skirts in the photo. The computer found four skits, which is correct. Now, if they can train it to pick up on colors, they’ve really got something.

This tool is great for marketers in two ways. First, it has the ability to turn Pinterest into a first class image search engine. More functionality means more users. More users is good for any business who maintains a Pinterest account.

On the backend, Pinterest can use this tool to locate and count like items then use that to sell ads. That means they can prove that more that 50% of users upload and share party dresses or funky shoes or designer handbags. And if they can locate those images, they can target those users with appropriate advertising.

Right now, Pinterest is a toy. With a sophisticated image recognition system, they’ll go from toy to tool – and I mean that in a good way.