AI location from photos

[GoJeff!]

Not sure if this is old news, but a friend sent me this link of an AI beta to find the location from a (non-metadata) photo.

It was able to locate a childhood photo with very little of interest in frame to within a few miles of my house. The ones I've tried since have been less impressive, but it occasionally nails something that seems impossible. Kind of fun.

https://geospy.ai/

 

[Kenny F'ing Powers]

Oh boy. With people posting everything they do to social media, I can't possibly see how this ends poorly for some poor girl.

 

[NJ_Sox_Fan]

I did a few of these and it got every single one very wrong

 

[jayhoz]

I tried a couple and the descriptions of what the tool used to make its guess are even off. For example, it told me there was a city skyline in the background when there was nothing but sky.

 

[Lose Remerswaal]

I did a few of these and it got every single one very wrong

to be fair, one highway in New Jersey kind of looks like every other highway in New Jersey

 

[hoothehoo]

I just submitted a picture of my local library with a sign out front stating it’s name

The website correctly read the name but placed it in a completely different state let alone town

 

[jayhoz]

I just submitted a picture of my local library with a sign out front stating it’s name

The website correctly read the name but placed it in a completely different state let alone town

I didn't strip any of the exif data from my photos so the thing isn't even smart enough to cheat.

 

[SumnerH]

It did okay with mine, though a lot of the details are wrong. 15 miles and a couple towns away, but right metro area:

Country: United States​

State: North Carolina​

City: Asheville​

Explanation: This is a residential street in Asheville, North Carolina. The trees are deciduous and the leaves are green, which suggests that the photo was taken in the summer. The houses are mostly made of wood and are of a similar size and style, which suggests that they were built around the same time. The neighborhood is located in a valley, and there are mountains in the background.​

Estimated Coordinates: 35.5955° N, 82.5512° W​

 

[saintnick912]

This seems like it would be a precursor to a Geoguessr bot-solver.

When I shoot on a camera without GPS tagging (my latest has it), I tend to take phone pictures in any non-obvious locations so I can go back and correlate them later. Sometimes I have to interpolate between the location of two phone pictures to get a bearing, but I end up tagging locations on just about everything in my Flickr that I don't want to intentionally omit.