I’ve been hearing a fair bit lately about how amazing Google Photos is at intelligently identifying the subjects of photos. On numerous podcasts I’ve heard about people searching for “hugs” and “dogs” and “patios” and getting “astonishing” results. So I thought I’d give it a go.

I first installed the Google Photos app on my iPhone and gave it permission to back up my Photos library from there. It set to uploading immediately and I gave it plenty of leeway to get into the task, but soon realised that was a bad idea because I’d have to leave that one app open on my phone for a long time.

Next I installed the uploader app on my Mac and gave it permission (only) to the (same, iCloud) Photos library and left it to its devices. Within a day all of my 5,000+ photos were in the cloud. Next up was to try out this clever searching.

First up – “trains.” About 80 photos appeared including very obvious photos of trains, but also photos of parts of trains (e.g. windows) and the inside of trains. Pretty clever. I figure these are all iPhone photos and therefore geotagged and all of those internal shots would geolocate to a railway line. Fair enough. Seven shots are of railway lines but do not include trains and two shots are of trucks and cars, in a carpark miles from any railway. Not a bad score.

Next up is “cats.” This is most impressive. About 300 shots and every single one features at least one cat. That includes a photo of a picture of an ornamental cat and a tiny face peering through a window.

Third in the list is “aircraft.” One shot inside a train, one suburban sunset, one of an airport apron with no aircraft (but a fire engine), a marina, a bus, and a snowy garden pollute an otherwise impressively accurate selection of a few hundred. This includes aircraft interior shots and out-the-window shots as well.

My fourth and final test is “cars.” About 58 of 300 shots include some kind of vehicle I would accept is a “car,” although this includes trucks, aircraft tugs, and a large crane. What’s more impressive is the range of other subjects that were selected. Brace yourself!

Aircraft (many!), a harbour sunrise, other harbour vistas, boats and ships, architecture, a kitchen cupboard’s contents, a train ticket, bicycles, a coffee machine, a tangle of cords, trains (including a few that the “trains” search did not select), airports, yachts, a store candy display, a footpath sign advertising pies, birds, creepy sculptures, a busker, selfies, a cat, a half built gazebo, and a house! And it missed some fairly obvious car shots, too.

So a bit of a fail on that last one! I’ll admit the first three were quite impressive, but as many people have said of other services  recently (notably Siri), if you can’t rely on it then it’s of considerably less value. My guess is they’ve taught the service how to recognise certain subjects, cars not amongst them.

But the one disturbing facet of the service, and the reason why I’m now removing all of my photos and the software, is that I have spotted a lot of photos which have been deleted from my Photos library. They do not appear in the Photos app, including in the Recently Deleted folder, and from memory, most haven’t been in my library in a long time. This gives me concern on two levels – first that Apple isn’t physically deleting the photos and second that Google are likely not using official APIs if they’re finding deleted photos!

Categories: Tech


Leave a Reply

Your email address will not be published. Required fields are marked *