Archive 1997–2024

CES 2009 - Day 1: Netbooks, "Talking" Cars and more

deepViewer: Picture Catalog & Search Engine

There’s this ridiculously smart Intel guy I meet with on a regular basis named Francois. He’s a big car guy, currently favoring the Nissan GTR, and also happens to be incredible knowledgeable about all things SSE and video related. For the past couple of years Francois has been showing me a little application called deepViewer pretty much every time I meet with him.

deepViewer is a photo organizing application, it allows you to quickly navigate through all of your photos. The pictures are organized in a large calendar, you can zoom in and out very quickly (assuming you have a fast hard drive, or SSD, and fast CPU); the whole thing is very movie like.

I’ve never written about deepViewer before because the application isn’t out yet (although it will be soon). While navigating through your pictures quickly is a cool feature, it’s not enough to get excited about. Today in my Intel meeting Francois showed me something a little more interesting.

deepViewer got an upgrade, it can now find all similar pictures in your library. Simply right click on your photo and then tell the software to find similar images.


The circles in the image above are the portions of the image deepViewer looks to match in its database. The larger the circle, the higher the weight of that part of the image.

Internally, the application samples a number of points from the image and searches for similar elements in all other images. The search works according to elements of the image like faces or backgrounds and colors. The search isn’t exactly live, all of this data is sampled when you index your images and it’s simply updated as you add more photos to your library.

The process of searching isn’t very CPU intensive, it’s the indexing that needs a fast CPU. Naturally, Francois demonstrated deepViewer on a Core i7 (it was overclocked to 4.6GHz using water).

Originally published at https://www.anandtech.com/show/2704 (Wayback Machine).