A little over two weeks ago my favourite search engine (favourite because I’m too lazy to try the others) rolled out voice search and search by image. I tried them out so that, if you’ve not yet got a chance to use them, you can raise or lower your expectations accordingly 🙂
The idea for search by image is simple: drag and drop an image into the search box. I used my own photos which had the default name and no extra meta information (such as geotagging, etc) that would make the search engine’s job easier.
Architecture
It got these two right:
It was able to recognise the Spanish Steps even with the people, scaffolding and temporary decorative flowers:
It got stuck on the 21st century Ara Pacis museum:
Landscapes
It got the idea of sky and water:
And that of sky and (mountainous) land:
Objects
If it can’t make out what’s in the photo, it goes for matching colours:
Faces
A human face is recognised as such, but I’d hope for more similarity (Picasa’s face recognition spoiled me):
Not wishing to use pics of my friends and acquaintances, I used my cats. Bolla looks like food (which is kind of funny considering Bolla eats A LOT):
Missing and presumed dead Tempesta is inscrutable, but at least there’s one cat in the suggestions:
A black and white shot of Wiki brings up lots of human faces, as well as an unexpected symbol:
Conclusion
I’d hoped that it would work better than it did, especially with faces. I guess that will improve with time.
Next: Should people with non-American accents bother with voice search?