Locating a Photo of a Vehicle in 30 Seconds with GeoSpy
geospy.ai
geospy.ai
Yeah I think we've figured out what inspired him to build the tool. :/
GeoSpy told me that it was the place mentioned in the picture, with the textual description as evidence.
To mock their user reviews...
> “Graylark helped me find the person I'm stalking in under 20 minutes. This tool is unbelievable — a true game-changer for those with restraining orders like me who just want to get back at them for that court order."
This mentality is kind of dumb, no offense. We have a bunch of laws. You could just as easily use your argument to say murder should be legal, or rape, and certainly people have.
Laws do, actually, work, for the most part. No they're not perfect, but they don't need to be.
Criminalizing everything that could be used to do bad things is an extreme position. Instead of jumping right to "ban it" you should probably first have a discussion where you consider whether (A) that ban will make any difference to its availability to most people who are criminally-minded anyway and (B) whether it has positive benefits to the law-abiding.
"No-one charged in 9 out of 10 crimes" https://www.bbc.com/news/uk-44884113
Laws are basically codified morals, but shitty because they need to be written to be some semblance of objectivity. You typically get stupid results when you try and surgically codify niche things or try and legislate controversial things.
I'd much rather live in a world with LLM image location stalking than one where people just punt everything to the state.
2. Why outlaw bombs if criminals have obtained them anyway? You’re just arguing against he concept of laws at this point.
3. A type of app is not synonymous with “an activity”
There's potential for far more, and far more lucrative corporate and state harassment here. Think like low effort red light camera mail ticket but for the general case.
"We see that someone has posted a picture of X at your location. Here is a copy. This is a violation of a) your leas b) the zoning code, please pay us $1000, if you would like to appeal please fill out the attached form and include the $500 appeal fee and if you lose the fine will be $2000. Reminder: you agreed to this in subsection ABC of <your lease|the zoning code>"
Who do you think is "sponsoring" this ? /s
I'm not diminishing the ethics debate, but it's crazy to me how easy it was for two non-technical rich dudes in a garage to build Clearview AI (And before vibe-coding!):
1. scrape billions of faces from the internet
2. `git clone` any off the shelf facial-recognition repo
While they are pictures on the internet it is one thing, when you gather them all and put a label with a number then it is problematic.
Remember that FaceApp to make you older, younger etc? Imagine how much data those guys collected?
I know someone who submitted the face of a member of my family without consent. You could not even complain without agreeing with the TOS first
It wouldn't be hard to narrow things down:
Year/make/model/trim/color/region/timeliness will narrow down to a very small if not already unique subset of vehicles. And on top of that vehicles may often have unique stickers, accessories, or scratches which can further strengthen the case. Flock e.g. uses this data in their vehicle identification algo.
The idea here is that you find a picture of your vehicle that the thief took, and use this to find the location of where the thief has your vehicle.
I thought you uploaded a picture you already had, it does the scanning, and a hit might look like “some rando posted a selfie at Zilker Park 20 minutes ago on insta and that car was in the background”.
Although your example may be quite viable in a repossession scenario where the possessor is known but the location is not.
This part still is the sticking point;
> When browsing Craigslist, I came across a regular car listing that showed a vehicle with buildings visible in the background. The listing claimed the vehicle was located in San Francisco. ...... Superbolt returned precise latitude and longitude coordinates that, when entered into Google Maps, revealed an exact match to the buildings visible in the listing photos.
How often do people find their stolen vehicles posted on CL/marketplace? Do police have resources to constantly browse hoping they see a similar picture of their stolen vehicles? How do they match it to the one they are looking for? Eg. if this was a cop, they may think, this vehicle matches the description of the stolen car. And this AI tells me the picture was taken at these exact coordinates (not super useful as this looks like a public place and I'm sure not where the vehicle is being stored). They still have to go out, meet the "seller", check the VIN or otherwise confirm it is the correct stolen vehicle they are looking for, then they get an arrest and recovery.
But, what if there are a dozen vehicles for sale matching said description. They now have to arrange to visit them all until they find the match or exhaust their options. How is this AI adding any value given with & without it the process looks the same; find listing, ask "seller" to meet, meet, evaluate. You don't need this AI to ask the "seller" to meet up and pretend to be an interested buyer.
FWIW, this looks like it could be a white VW Jetta to me. There are 118 in SF bay area right now just on Autotrader (granted, the hatchback is a further narrowing feature, but that's not super common either). No police department I've ever heard of has the resources to check on all these listings. If the thief stole it in SF but listed it for sale in Seattle or LA or anywhere else, how would anyone know? That's the haystack part, it's a big haystack.
> They still have to go out, meet the "seller", check the VIN
You do not have to meet the seller to check the VIN of a vehicle sitting on the street.
> But, what if there are a dozen vehicles for sale matching said description.
There might be a few vehicles in an area matching make/model/year. But it is trivial when looking at a photo to filter on further criteria... and once you look at the photo you can observe trim, exterior color, interior color, stickers, inspection sticker, etc, you will have a very high degree of certainty even on a common model.
> white VW Jetta to me. There are 118 in SF bay area right now just on Autotrader
Well yeah, because you only filtered on 2 of the dozen or so attributes that you might know.
Within a whole 500 mi of the bay area there are only 5 white VW Jetta Wagons listed. All you need to know is what year it is, to narrow it down to 2 or 3. If you know the trim, approximate mileage, any visually distinctive feature, etc, you are guaranteed a match. Even if it wasn't a wagon, it is not hard to filter down to a unique vehicle.
You're assuming the vehicle remains where the photo was taken.
> Well yeah, because you only filtered on 2 of the dozen or so attributes that you might know.
Those were the only attributes that were apparent in the photo. I said I ignored Wagon because that was a cherry picked unique filter. If it wasn't a wagon, your analysis is the same as mine, >100 vehicles in the SF bay area (I only filtered on 100 mile radius). But again, why steal a vehicle and post it for sale in the same city you stole it from? Criminals already move stolen vehicles, this is all but obvious.
Basically, this helps you catch the dumbest of the dumb criminals. Someone that steals a very unique car and posts it for sale in the same area they stole it from and also leaves the car parked in the same place they took the photo. There's also a time element, if they hide the vehicle for a few weeks, then post it for sale it's more likely the initial active investigation has faded and the cops aren't actively hitting refresh on marketplace.
Glad you believe this is useful, I'll continue to disagree - it might have some value but it's usefulness is being exaggerated in the article.
Casual small time occassional car thieves might do this, receivers of stolen cars as payment for other debts owed by a thief may do this ... but it's somewhat atypical.
> Aren't most stolen vehicles disassembled (chop shops, etc)?
In the organised bigger scale operations vehicles are dealt with for the greatest profit with least risk. A good many are stripped for the parts - the more popular the car, the larger the parts after market.
A suprising number of cars from developed countries are shunted whole into containers and sold elsewhere about the globe. eg:-
“Each year, hundreds of thousands of vehicles are stolen around the world, yet the initial theft is often only the beginning of a vehicle’s journey into the global criminal underworld.
“Stolen vehicles are trafficked across the globe, traded for drugs and other illicit commodities, enriching organized crime groups and even terrorists.
Local police are doing none of this btw.
Really? Not that anyone has any data on any of this but since you're measuring it as "often" I'm going to disagree and say this is a very tiny percentage of stolen vehicles that are being used this way.
If they are, it's probably being bought from a hookup you know and not randomly on marketplace.
geospy.ai: the real technology seems to be that they invented the world's thinnest veil
Clearview could be eye drops. GeoSpy: well...
They contrasted this with their prior technique which is more of an image classifier that can identify general location from image features. This approach does not require their database to contain a picture of the exact location.
First things first, its entirely possible to geolocate using just visual markers.
A bunch of startups did it around 2018 (most got bought by facebook, ie mapillary) They work by extracting keypoints from pictures and building a massive point cloud of identifiable key points.
But
That picture they use with supposed keypoint matching is wrong. None of those keypoints are reliable feature descriptors. They all are on foliage, which changes depending on season and wind. Geolocating that picture accurately _automatically_ using features is next to impossible.
Now, they might have a vibe based matcher which does some basic spatial comparison, but I'm not sure how reliable they are, especially given a large search radius.
The other interesting question is, where did they get their data from? I'm pretty sure google spent a lot of time making it really difficult to train from street view (lord knows we've tried.)
Edit the demo here: https://geospy.ai/ is much more what I recognise a bog standard VPS system does. Note that the user is matching buildings. Thats far more reliable way to do feature matching.
Seems plausible enough to me. The trees are evergreens in a place that doesn't get snow, and the keypoints are mostly grounded on stable parts of the trees (trunks or thick branches), which barring gale-force winds probably don't fluctuate all that much.
The part that gives me pause are the keypoints that map the hood of the car to the pavement, and the point on the far right that maps the ledge to the pavement. How can a system robust enough to map foliage also return such blatant false matches?
long answer, have a try on this demo: https://docs.opencv.org/4.x/dc/dc3/tutorial_py_matcher.html
short answer is that they are similar enough features to match. think of them as homophones (ie words that sound the same but have different meanings) in language. You need context to be able to filter them out. (https://github.com/polygon-software/python-visual-odometry/b...)
> don't fluctuate all that much.
Over time that doesn't bear out. Good features are areas of high contrast with nice clearly defined edges (text is great, so are buildings). branches move, which means they create lots of diffrent features depending on the wind, even light wind. when we were building out maps, we filtered as much greenery out as possible
Why America’s Heroes Deserve the Most Advanced AI
our goal was to build technology to safeguard American freedom and prosperity...
...America deserves more. While Silicon Valley hype centers around LLMs, AGI, and SSI, our focus remains on visual intelligence—understanding the world we see with our eyes, what we call Visual Super Intelligence
https://geospy.ai/blog/why-america-s-heroes-deserve-the-visu...
Services like this (Flock, etc.) should either be illegal or accessible to everyone.
I guess I grew up drinking the 'American culture is one of mistrust of government' cool-aide, rather than 'American government has deep pockets' fruit punch.
I'm not sure if it's just an evolution of the times, or an actual erosion of principals (since when? 9/11?)
> misspells "Kool-Aid"
Okay then, thinks EB, mentally trying to decide which photos to try it with.
"Look here's a picture of a place, and here's a pin on a map that shows you where it is!"
Yeah, I can do that without AI.
A repo investigator for the bank locates the target vehicle via owner’s social media, takes photo of the car, shoots it into GeoSpy, then ganks the car based on given locations in the owner’s photos. Pair it up with ALPR hits across a city from national ALPR networks (to help correlate home/business/work patterns) and… wellp, there you go!
Flock is an ALPR.
most commonly it is used for drugs in canada since every case I hear about ends up in forensics
The stryker will find or buy from somewhere, a pool of unissued VINs that don't flag anything in the state registration system and match various vehicles (Dodge Chargers, Kias, Hyundais). Then when someone comes with that vehicle, they will strike a new Vin plate. Sometimes if they buy the VINs it will come in a package with plates. From there it's possible to get the vehicle registered, most likely under someone else's name that has no idea and they will sell / rent the car with tag etc. Though sometimes they will just make a fake plate too and then steal a real plate, swap it with the fake plate and put the real plate on the stolen car and sell it like that. In some cases / states they can actually get a title reissued.
Boats are even better, but much smaller market, just look up coast guard plates on Amazon.
Stiker vs Styker, is regional.
For Reference: Striker Music https://www.youtube.com/watch?v=AaTxkD5JFpg
Who else exports their car, doesn’t report it and then offers their VIN?
So, those seem like they'd be pretty good ones to use, as the straw buyer would certainly not report it as stolen. Though I bet they don't renew the tags, so you might owe a couple years of back registration depending on how old the 'source' car is.
They get all cranky about people arbitraging it but it is blatant price discrimination.
Manufacturers were starting to require proof of insurance before handing over the keys and then people would get it and cancel+refund the insurance. Cat meet mouse.
https://www.cbc.ca/news/canada/british-columbia/icbc-luxury-...
It sounds like this scam would only get discovered when you go to the dealer for service, perhaps.
The car was sold on, eventually went to Copart with a blown engine and then the YouTuber found out through his videos that the car he owned was stolen and the original had been exported because the interior color was not the same as the decoded VIN. Only when he took the engine out of the car and compared the engine number with the one in BMW's database and the reported VIN in the infotainment was he confident that the car was stolen, same for Copart (who wouldn't entertain the car was stolen).
I think if it wasn't a famous YouTuber who bought the car, it's highly possible that the stolen car would go nu-noticed throughout its lifetime as stolen, even if taken to a main dealer. If I recall correctly the car reports he used (maybe car-vertical) also didn't pick up any discrepancy.
For the criminals its good business, you find a 30k plus car, pay for a clean VIN from cypress or somewhere and then do the damage to the car to re-new it as a different car, even if it costs 10k to do, its a lucrative 20k "profit" and thats on the high end, seems like cars can be stolen overnight, especially ones the criminals specialize in.
But thieves don't really really care about what it technically represents, they are more interested in what they can get away with. That would be solely dependent on how stringent the inspection is to get a rebuilt title.
"law enforcement agencies can achieve faster resolutions, greater efficiency, and better outcomes for vehicle theft cases"
Could and would are two very different things in America.
In most cases, the police would simply do nothing.
Facial recognition technology (see Facebook auto-detecting your friends when uploading a photo) has existed for decades. Why do the police still post photos of suspects asking the public in help identifying so-and-so?* Can't they cross-reference with the DMV database or even Facebook to see if there are any matches?
*Although these days they even stopped doing that, I've seen cases where they blurred out the suspects face and then asked the public in help identifying them. They do this to protect the criminal's identity. Sigh. I wish we could bring back name and shame.
I think there are laws that bar them from doing that.
Thanks for the post, AMA for anyone into computer vision or AI.
:)
That seems like a stretch. That wouldn't even make sense for them to do. Strange claim to make.
>As of Oct. 15 (2024), the DMV reports that it has recovered 228 stolen vehicles amounting to a value of $6.35 million. Of the 228 vehicles recovered this year, 149 were purchased by an unsuspecting victim after seeing a post on Facebook Marketplace or a similar site.
That is 65% of recovered stolen vehicles in New York was sold through online marked places.
https://www.silive.com/crime-safety/2024/10/stolen-vehicles-...
1) How many total stolen vehicles there were
And
2) If 65% of recovered ones being from Marketplace means only the low-hanging fruit were found.
From https://apps.criminaljustice.ny.gov/crimnet/docs/FINAL%20202...
PSA for anyone with a bike: register it on https://bikeindex.org/. I registered mine there the day it was stolen and almost immediately got emails from several people who were monitoring a known bike thief's Marketplace account and saw mine listed.
That way they only take the risk when they need to, they don't need any storage area, and if they are caught it is only for 1 bike, not tens or hundreds.
https://www.youtube.com/watch?v=Vo0gualrJa4
Go to 5:30 for a taste
"It's used for car theft!" except the intended use is obviously target government buyers for tracking citizens.
Later edit: Something like this [1], from late 2020: "Russian Vehicle Registration Leak Reveals Additional GRU Hackers". There were also some other articles with Google StreetView screenshots and the like, I won't search for them because even finding this one reference means I'm not doing something better with my time right now.
[1] https://www.bellingcat.com/news/americas/2020/10/22/russian-...