When machine learning helps us find photo memories

I am still amazed at how smart machines are when it comes to understanding what we include in our photos. I just ran two simple searches on Facebook and Google+ Photos of my photos and received these results:

I think Google’s machine learning is better when it comes to semantic searches although I haven’t conducted any scientific tests of any sort. It doesn’t really matter which is better. What does blow my hair back is that you can search for objects in the photos and have these machines show you those photos, even though your titles, tags or other metadata has nothing to do with your search terms.

I’ve been thinking about the best place to share my photos and I am very tempted to stop using Flickr and, to a lesser degree, 500px to showcase my photos and to use Google+ Photos instead. As a social network, Google+ hasn’t exactly made waves but it is an incredibly dynamic and powerful photo sharing service.

One of the aspects of Google+ Photos that I really like is Stories (here is a fascinating article about what goes into making Stories possible). Rick Klau’s post about how he started adding old photos to Google+ Photos with some pleasantly surprising results got me thinking about this again:

I prefer to edit my photos myself but the Stories feature in Google+ Photos can be a really nice way to share, well, a story that I capture in my photos. When you add the machines’ ability to recognise things in the photos and make them so much easier to find (or even discovery forgotten gems), these sorts of services become really compelling photo sharing services.

Wouldn’t it be interesting if Facebook introduced something like Google+ Photos’ Stories to its photo experience. The combination of the datasets these services have with that sort of nascent intelligence can be remarkable.

On a related note, Om Malik has published a very interesting essay titled “On visual web, a photo is worth more than a 1000 words” which is worth reading.