Checking on numberplates and bin logos was hardly groundbreaking, as those are exactly where I would have started too. However, I had never heard of Overpass Turbo [0]. Learning of its existence was the only valuable take from this video for me, but well worth it.
Basically, that is the gist of the vid, the rest is more or less common sense, I guess.
A short & catchy variation of "Overpass turbo - a super powerful mining tool for OSM" would be a more appropriate title [0][1].
It exists since 2013 but got a massive boost from the Pokémon Go craze in 2016 [2].
And in this case I think it would have been a lot of work to identify the number plate & bin logo if he hadn’t already had enough context to make a probable guess. The source image he was working from looked pretty low-res, that license plate was just a couple white-ish pixels. Like you, I hadn’t heard of overpass turbo and that looks super cool, but the rest of it was “how to confirm what you already mostly know”.
I mean, it's probably niche if you consider all software ever, but the tool has been around since 2013 at least (and before that under a different name, OSM Server Side Scripting/OSM3S), with a lot of ecosystem tools and documentation (in multiple languages) existing all over the place, I wouldn't say it's very niche in the mapping community.
I agree, GPT3.5 has been working insanely well for me when making overpass turbo queries. I guess there's lots of training data in the dataset for it. Made a quick 10 liner that takes in a bbox & search prompt -> GPT3.5 -> overpass turbo -> geojson
In my experience it seems like prompting for older technologies works much better on large LLMs, i guess it makes sense considering that there is probably more crawlable documentation out there
It's a surprising helpful tool in Pokemon Go metagaming. For example, you can use it to determine gyms eligible for EX Raid status (not that that's a thing anymore), find potential micro-nests, or even locate Lake Trio spawn locations.
One of these I had seen had almost no hints, but it did have a church that was fairly close to a kfc and a dollar general or something. It also kind of “looked” like the southeast, so the dude did a general query on places where those three types of structures inhabited the same general area. It came back with about 7 or 8 results in the entire southeast which could then be manually checked so it seemed like being able to query the map did the big lifting there.
I think the real discussion here should be an analysis of his methods of using the various online tools... his toolchain, essentially, and how we're rapidly accelerating processes like these compared to even just 10 years ago.
Openstreetmap data is used in Overpass Turbo -> ask ChatGPT NLP into Overpass Turbo scripted API -> copy paste script into Overpass Turbo -> select Location data from query -> copy/paste coords into Google Street View images -> Matching the original image by eye.
This part of the his search could be collapsed into a single tool. Or, if that is maybe too much of a specific use case, imagine moreso a 'workbench' tool that would put all this in a single place. Every time he switches chrome tabs, or copy/pastes... that could be taking place in a single environment, like a GPT4-plugin for example.
I wonder how hard it would be to develop/train a machine learning method that ingests a photo of anywhere outside and maps it to corresponding satellite/aerial imagery.
We are of course very good at object recognition from still images. We are reasonably good at inferring depth from a single still image (monocular depth estimation [0-2]).
Combining these, we should be reasonably good at determining relative positioning of identified objects in 3D, and thus computing their overhead 2D positions. Imagine discretizing overhead x/y positions, essentially yielding an image where each pixel corresponds to a 1 meter square looking down, whose color is the inferred identity of what occupies it.
For example, in the linked album cover, we could infer that a street occupies a set of pixels with coordinates S = {(x1, y1),...(xN, yN)}; a building occupies pixel set B1 = {(xb1, yb1),...(xbM, ybM)}; another building pixel set B2; etc.
Now, we are also reasonably good at object recognition (and MDE) for overhead imagery [3]. This would let us build a giant set of overhead object identities inferred from the entirety of satellite imagery. The tricky part would be efficiently indexing this, such that the overhead coordinate sets inferred from the ground-level image could be quickly (and fuzzily) queried in the satellite data. A brute-force approach would essentially convolve the overhead object coordinates inferred from the ground-level image over every region of the satellite imagery, at every possible angular orientation (and likely several different scales, since MDE is often imperfect at inferring absolute depth), but this would be impractically slow.
I would bet that intelligence agencies have had this capability for some time now.
note that when they went and analyzed that AI they found out it was using the smudges on the google car camera as a sort of fingerprint (as those are consistent for the many pictures taken by one camera) so that AI would almost certainly not do well against pictures taken from different sources
>The model [uses] semantic geocells – bounded areas of land, similar to counties or provinces, that consider region-specific details like road markings, infrastructure quality, and street signs
>The model can['t] pinpoint exactly where a street-level photo was taken; it can instead reliably figure out the country, and make a good guess, within 15 miles of the correct location, a lot of the time
I could imagine that a more qualitative model like this one could be used to restrict the search space of a more quantitative model akin to the one I (very roughly) proposed.
I think Google Maps is doing something similar where it sometimes asks you to take photos/a video of the street you’re walking on to detrmine your location more precisely (I think it happens when the GPS is not giving good enough precision).
There's a separate augmented reality mode which a) captures frames in real time to pinpoint your position and b) displays the video stream with overlaid 3D markings which show you the way.
Maybe this is common knowledge in the US, but like, you're just supposed to know what Kentucky number plates look like from a blurry image, and you're also supposed to know the logo of a sports club?
Cool.
Also, picture has to contain all these clues of course, conveniently, including a couple of street numbers.
> you're just supposed to know what Kentucky number plates look like from a blurry image, and you're also supposed to know the logo of a sports club? Cool.
Knowing exactly what they look like would help, but the point is more that you're supposed to look them up and see if they match to rule out regions or confirm suspicions. You've got the wealth of human knowledge at your fingertips, use it - if knowing what a Kentucky numberplate looks like might be useful, you can just look that up in a matter of seconds.
> Also, picture has to contain all these clues of course, conveniently, including a couple of street numbers.
Given how many times this individual has narrowed down an exact location from a photo of a straight road with nothing but trees on either side, I think it's fair to suggest that he was picking an easy target for the purposes of tutorialisation.
It might take more than a couple of minutes to narrow down a photo with less obvious clues than this one, but it's fairly difficult to take a photograph outdoors that doesn't give away enough for someone with sufficient knowledge to identify exactly where it was taken.
The main clue is knowing that Jack Harlow is from Louisville Kentucky, so he is basically guessing that the picture was also taken there and just confirming it by checking if the license plates and logos match those he can find for that area.
I wish he'd have done a harder one. This one felt too easy, because you can just Google where he's from, and just Google search '1948 address louisville ky'.
overpass turbo is a tool I didn't know about, which does some of the heavy lifting here -- and could do even more in other cases, his example of "skate park within 100 meters of a dog park" makes me think. i didn't actually realize there were public databases available with that kind of granularity, coverage, and coding.
The ability to have ChatGTP generate the query is actually pretty huge.
I wonder how long until there are AI tools where you can just say things like "give me skate parks within 100 meters of dog parks in Boulder CO", and it will identify the tool to use and generate the query and execute the query and just give you the answers.
That public database is Open Street Maps which I’d describe as “Wikipedia for map data”.
It’s amazing plus it’s probably used in several tools/apps/services you use regularly.
And best of all, if you have a look at the map around someplace you’re familiar with there are probably several improvements and additions you can make this very day using their online map editor.
It’s a meme now on his channel that he works for the CIA given his recent travels et al.
He made a funny video about taking the “test” on the CIA’s website which was an online game for kids (a really weird thing to exist when I stop and think about it).
I once saw a Johnny Long talk about how to do this for interior photos, where something like an electrical outlet or carton of milk can make all the difference in the world.
Switzerland has a different electrical outlet to the rest of Europe (Type J), and an easy way to spot fake apartment listings is when the pictures of the apparent apartment show "EU" outlets (Type F).
I have a lot of old pictures without geolocation data. A tool to automatically add even an estimate of the location would be really useful so that apple or google photos could use it to make albums automatically.
Getting http 403 errors in NewPipe after the first minute of watching. After twitter and reddit, I guess it's time to wall another garden :( wish people would at least cross-post information to literally any other website (so there's a semblance of market forces—reasons to be nicer to use than the competition—and not just 1 place you have to use)
For their entire existence, newpipe and youtube-dl/yt-dlp have frequently stopped working and needed to be updated to deal with changes in youtube. This is nothing new; youtube was never open.
It's also not as if most other video sites provide an open api that doesn't need to be scraped
Genuine question, what exactly is the security problem here?
It’s a set of basic techniques to find where a photo was taken. I have a hard time envisioning an actual new issue here.
Obviously, people shouldn’t post pictures of things they don’t want other people to see or know (like where you live if you don’t want people to know that) but that was already true before people could easily mine Openstreet map data.
Overpass is not just "mining OSM data", overpass is quite an extremely efficient language and system plugged to a GIS database, that can be used to find a lot of things.
Those things could be used by robbers or any other kind of people who browse photos online.
Now, of course it depends on people handling their data straight, but since the internet is mostly used as a place where people share anything to the public, I still believe it's a bad combination of problems waiting to happen.
> Overpass is not just "mining OSM data", overpass is quite an extremely efficient language and system plugged to a GIS database, that can be used to find a lot of things.
I think I’m lost here because what you are describing is exactly the meaning of mining openstreet map data. That’s the GIS database they use.
But the barrier to entry to being an effective bad actor is much higher, and some harm is prevented even if not all.
I think your point is valid to consider but it uncharitably overlooks the positives of the other side's proposition and reduces it down to all or nothing.
Haha he's a pretty entertaining guy. Good tip on Overpass Turbo. It's wild that open data is full of so much stuff and enables all these use cases. Big fan of the tool as well.
I genuinely can't work out what part of this video you think is US-specific. America is far from the only country in the world that has municipality logos on infrastructure, house numbers, or OSM coverage.
There's certainly regions of the world where doing this would be much more challenging (e.g., Central Africa, China, rural India), but the stuff he covered in the video is going to be extremely helpful in the vast majority of cases.
One thing I noticed that doesn't generalize is exactly those house numbers. Depending on house numbering scheme, a house number may effectively be useless. Here in Israel house numbers are just sequential and most streets aren't that long, so what are you going to do with a number like "17" which appears in almost every street in the country?
> so what are you going to do with a number like "17" which appears in almost every street in the country?
Exactly the same thing that Rainbolt did in this video: cut down the amount of work you have to do on each street from checking dozens or potentially hundreds of photos/angles to just 1-3.
Of course if you've only narrowed the streets down to 20,000 candidates instead of 20 of them, that doesn't get you straight to the answer, but it's still a massive proportional improvement.
But the lesson being presented here is to use data that's available to you in the photograph. Maybe you don't have any street numbers (or any particularly useful ones) visible, but you can see the sun at the end of the road and therefore know that it's running directly East-West. That filters out tons of roads. Maybe you can see that houses are only on one side and a river is on the other, you can use that as well. In the video he mentions similar constraints with regards to local parks as being other options for this kind of search narrowing.
The point of that portion of the video isn't "hope the house number is really weird lmao", but to extract any geographical information out of the image and then query that with open mapping databases. House numbers are just one of the most common ones, and while they're typically not quite as powerful as was shown in this video, they're very often going to help dramatically.
Same here, and to that I don't think I know of any municipality logos, pretty sure that is not a thing. And all license plates in the country are going to be the same.
There's a few open data projects that provide similar features to Streetview and can very occasionally be useful in regions without coverage. But yes, typically outside of Streetview coverage you'll be relying on satellite photography, which makes things far more difficult. But that doesn't exactly make the advice here any worse, it just means that the problem you're trying to solve is fundamentally more challenging, so even good techniques might not be able to get you to the solution.
I haven't used his method (or tools) for finding locations. On his video though he does mention Austria/Vienna for museums, etc (on his screenshare), so I assume (assumption1) that if it shares data with OsmAnd (and others) it would be very useful for (anywhere) where internet is prevalent (Americas, Europe, Oceania, most of Asia).
> The purpose of the Korean Land Survey Act was to prevent anyone who is threat to national security from stealing the country’s maps during the post-war period. ROK government enacted the law in 1961 and it concerned the establishment and management of spatial data — it has subsequently been amended in 2009. The Article 16 of the act states strict regulation on taking maps, photos, the results of a survey, or any land surveillance data abroad, because of the likelihood that it could harm South Korean national security interests.
[0] https://overpass-turbo.eu/