84 comments
Related:
My toddler loves planes, so I built her a radar (November 2023) https://news.ycombinator.com/item?id=38435908
My toddler still loves planes, so I upgraded her radar (January 2024) https://news.ycombinator.com/item?id=39094288
That was such a good read. Thank you for sharing.
Watching football this fall, I've been thinking about a little box that sits on my table and mutes the TV whenever an ad comes on.
Not decoding HDMI HDCP or anything, but a webcam + AI or whatever that watches football with me and mutes ads. Similar to plane detection, maybe. Are there any projects like this?
Nice idea!
Related:
In Black Mirror, they pause the advertisement when you disable the sound.
So maybe you'll eventually have to use a separate sound system whereby they cannot detect that you're muting it.
Then, they'll track your eyes and require you to watch the screen :)
At which point do you stop consuming?
> Then, they'll track your eyes and require you to watch the screen :)
Check out the recent controversy over Activision and Call of Duty. Evidently the game code requires access to your webcam and the company has patents on using your gameplay to train a bot that play like you do, then uses the idle time to make the bot play while you are away. This bulks up the pool of players ensuring you are matched with people resulting in a good experience for you. There's also a patent detecting your emotional state while you view their in game store, possibly for adjusting price to influence you to buy or spend more.
>Evidently the game code requires access to your webcam
Source?
>then uses the idle time to make the bot play while you are away
Surely they're better off using a GPU cluster to do training like every other AI company?
What if you don't have a webcam, you can't play the game? I'm skeptical.
In France the first TV program (TF1) now has unskippable ads. Until now you could just start later, get back to the beginning and fast forward the ads. I am not sure how this works if you recorded your program.
I do not watch TV but it was in the news.
Drink a verification can to continue.
Black mirror haha I think Samsung actually does it.
Yeah... of course our execs took that Black Mirror course. Don't want to get left behind.
"At which point do you stop consuming?"
The tech cultists won't, will always think "urrhurr let's counter this corporate restriction / government regulation with <stupid tech solution #921>.
It's so rewarding to do simpler things in life instead, f.e. growing your own food and not partake in this "more-and-more pervasive tech spread" BS.
At some point pirates come along with their own streams offering better service :P
MBAs think they can enshittfy into perpetuity, but eventually people have enough and leave to whatever the competition is.
i'm about to do it myself. the streaming services are just generally putting out shitty content (there's the occasional gem), charging more, and creating more and more walled gardens. it sucks to be a consumer right now other than the "ease" in which you can access these services.
I'd settle for a device that just normalized the volume levels, but this would be even better!
In the USA since 2010, there have been standards that prohibit ads being louder than the program material:
https://en.wikipedia.org/wiki/Commercial_Advertisement_Loudn...
Nvidia shield can do that, they call it "nighttime listening mode" or something, I leave it on permanently, much better experience.
Here’s a DIY project to make a TV volume limiter, I haven’t done it myself but it looks cool: https://www.youtube.com/watch?v=j1V2I-otdzk
Usually football trasmissions have identifiable elements in the screen, such as the score, the match title, etc... and those are fixed elements that do not appear in ads. For example: https://i.imgur.com/blebca8.png
Or the title bar below. You could do some basic OCR on a Pi, when the team names are readable, keep the volume up, when they are not, disable it.
Years ago Microsoft TV had a free 3rd party extension that analyzed the video and skipped the commercials. Amazingly you could see the downloaded video, there wasn't drm on that in the early days. There were several tools like this.
I think you could do this. If you had a web page logged into the same program it could analyze the video. I wonder if people build "video scraping" from web page tv apps, probably some drm blocking thing.
There also used to be subsonic chimes or something simila on the network tv audio to signal to some other server (maybe at the local tv station?) that would know to automatically run a commericial right now.
Looking at the electric consumption of my 30 closest neighbors using an RTL-SDR, I'm still wondering what I should do with the information.
only ethical move would be to move your own consumption to when theirs is lowest and not retain any data but your own
Are the electric readings sent in plaintext or trivially decrypted?
Unencrypted, I'm sure.
I hadn't thought to look for smart meters, but my RTL-SDR (and my Flipper Zero) can pick up all sorts of temperature sensors, etc.
A quick google found this rtlamr project with a number of blog posts about people doing this with it.
Perfect timing! I got annoyed by Ads in streaming services and decided to build something to mute the TV when an Ad comes on. Simple vision based system.
I'll post a write up in a few days.
Probably don't even need the webcam, detection via audio seems possible.
Rk3588's have HDMI input (I think with hdcp decoding) that you can use v4l2 API. The multiplane API can be a bit tricky, but otherwise somewhat trivial
I have had this exact same idea, it’s a good one. We all hate ads, and muting the audio is all we really need!
I wanted to do this too! I'm an "idea man" that doesn,t deliver, though. Here is my feature list:
* auto commercial block using audio fingerprinting and duration, so if it finds the ad, it knows the ad will last 10-15-30-60 seconds.
* frame-slowdown. often ads just show way too many frame jumps in rapid succession. If this is detected, visual mute would freeze one of the frames and only allow a jump after 5-10s.
* visual mute - if the commercial is not identified audibly, have a blur or black box that covers 85% of the center of the screen. you would un-mute it when you recognize your game from the borders.
* logo-blur. OpenCV to recognize most common ad logos and blur or black them
* commercial-over-dub. allow funny joke audio to be played over existing, known commercials. So when they show insurance ads, someone can say "yeah they didn't pay my $40k roof claim because I didn't have hail coverage", right over the insurance logo
* announcer over-dub. This would allow aspiring sports announcers (or comedians) to have their audio played in place of the game audio. This would require some visual sync, which would include a few seconds of delay. I bet there are some really good announcers out there who would love a chance to announce a game. Also some really funny smart-asses , or you could choose a biased announcer who roots with you for YOUR team and disparages the other!
how about fingerprinting? Not versed in this space but might be worth looking into.
Why not, HDCP is only a compliance scheme.
I recommend to not show dates as it can easily help triangulate your location of where this Pi is running, and with the internet crowd, doxxing yourself is never a good idea.
How much accuracy can you possibly get from a 600x600 image?
From the regular images, probably not a lot. But if you see a plane crossing in front of the moon or the sun, you could determine a ground track of the shadow of the plane. If you get that with two different planes, you probably have a pretty accurate position (a few hundred meters). Combined with the image of the setup, showing a balcony, someone could probably find the exact location.
> If you get that with two different planes, you probably have a pretty accurate position (a few hundred meters).
What?! I've seen the viral guy on social media who can pinpoint places in the world by looking at a photo but... a few hundred meters based on the angle of the sun/moon? That's wild!
If you have time of the image accurate to the minute, and you know that the plane, sun and observer were on a line, the observer has to be on the ground track of the shadow of the plane during this minute. If you have that from two planes on different tracks, the observer has to be at the intersection of the ground tracks.
This method is not just from measuring the angle of moon/sun, but using the plane (which has a known position) as a marker. Just by using the angle of sun and moon, you need to be extremely precise. 1 degree error is 1/360 of the circumference of earth distance, that’s about 100km.
If you know the flight and the date and time you can easily pinpoint exactly where the plane was and where it was heading. Not incredibly challenging at that point to narrow in on roughly where the photos were taken from.
Well we already know it roughly, it's Indira Gandhi International Airport in New Delhi.
>If you know the flight and the date and time you can easily pinpoint exactly where the plane was and where it was heading.
You'd still need reference data from the photo to triangulate where it was taken from. A blurry 600x600 photo is going to have plenty of uncertainty.
>Not incredibly challenging at that point to narrow in on roughly where the photos were taken from.
Most people probably don't care as long as it doesn't pinpoint a specific address.
Those red-green navigation lights are enough. There is only one aircraft position which would satisfy a given relative configuration. PI camera has a known field of view and given hundreds of images one could have directions converging pretty accurately.
Nit pick, it's not using a Pi camera, its using an old Nokia 5.4 phone as a camera.
Also, they're showing (and linking to) FlightRadar24's departure and destination data, it's quite clear they are on one of the approaches to Delhi International (DEL). I don't know how many runways DEL has, but there's no more than 4 possible lines planes take in approach and departure. If i cared enough I could easily work out which of those lines by checking FlightRadar for the history of one of the links, and match the timestamp with the FR tracklog.
Collection 100s of images any attempting to do geometry magic is unnecessarily complicating things. (Which is fine if your goal is to solve complicated problems for fun. But there is a much easier way.)
I'm under the impression that the OSINT crowd routinely comes to more impressive conclusions based on worse source material.
Depends on how far from the airport you are, and the accuracy of the timestamp. If you have accurate-to-the-second timestamps and you're within maybe 10km of the airport I think you could get down to a neighborhood pretty easily.
You know roughly the perspective that the camera sees the plane from, so you take the plane position at that timestamp and project that perspective line down onto the ground. The higher the plane, the more error there is with estimating the observation axis, so the less accurate this gets.
Maybe if the sky is cut away around the plane, I guess you could still make out a shadow on the tube/from the wings/engine nacelle
Really cool. I've considered doing something similar for alerting me of things seen in the air but without a transponder turned on. Good tip about the birds, I wouldn't have anticipated that ha.
Aside, the way air travel still happens out in the open in terms of communications data has a real early Internet vibe to me.
Very cool!
This plane suspiciously looks like a bird though: https://i.imgur.com/cUdnZTN.png
Haha, I'm working on fixing that. One of my ideas is to run inference again on an image to check if it has any birds. If yes, reject the image.
It's a really cool time to be checking right now! All commercial flights are grounded in Delhi for Republic Day celebrations so I'm able to see images of Air Force planes!
Could you get the speed and location of the plane and estimate how far it should move across two images to determine if its the right target?
That would work if inference on the Pi was faster. Right now it takes about 2.5s per image. The planes are in view for maybe 3s. By the time the next frame is fetched, the plane's already out of view.
This is so cool!!! Thanks for sharing the setup as well.
This is great. So I’ve thought of doing something slightly similar but with a raspberry Pi and camera to identify if there is a free parking space right outside our house (we live in a terraced house with no driveway)
I think this has been done for public places. A lazy approach (not using ML) is cropping/contour finding, then masking (opencv) against an empty parking lot, so if something is there like a car, would change the result.
The black parking lot would change in shade/hue so have to account for a range.
Thanks for the reply. Yeh I’m sure something like this exists. I could probably just mask out the area around the space o lab the streets outside our house and then send a notification when it becomes available.
The complexity maybe comes in in identifying our cars parked there vs someone else’s, so maybe I need number plate recognition. However my camera would be top down so likely no number plate would be visible.
Any thoughts/recommendations are helpful
Given that drone warfare is now a thing this seems like it could be useful as the basis of a personal drone detection and warning kit that soldiers could wear over their helmet to spot drones and give say an audible warning in the soldiers ears that lets the soldier locate the drone direction and maybe azimuth with pitch - lower frequency for a low drone and high for one more overhead. Not sure if such a thing already exists. That and a shotgun would be a reasonable defense.
Shotguns have proven ineffective as drone defenses. Range is too short and the targets are too fast.
Was thinking of some recent videos where a small drone closes right up to an individual soldier and explodes - assuming those aren't AI generated. But I expect there will be drone hunting drones and drone dogfights will be a thing.
Ukraine is using shotguns on drones to hunt other drones. Seems it works.
They are using those mostly to take down larger reconnaissance drones. Not the small ones used to directly attack individual soldiers. Watch the videos.
I never said it works against all drones. It is a valid option that has shown efficacy. In the context of this thread (one's home/property), it seems reasonably applicable, and in many locations, more likely to be legal than many alternatives.
Wrong. Shotguns haven't shown any real efficacy when used by soldiers (or homeowners) on the ground. In the USA it is a federal crime to shoot at aircraft, including drones.
I guess you can go tell the military they are wrong, along with all the homeowners that have been arrested for shooting drones down with shotguns. You might want to do some actual research before just spouting off whatever you think when the facts don't support it. They even make shot shells designed to take down drones.
Yes shooting down a drone is an offense in the US, unless you're in law enforcement or the military in specific situations.
https://www.armyrecognition.com/news/army-news/2024/analysis...
They are better than other guns though.
The set up: https://obviy.us/blog/pilane-spotting/
Super cool!!!!!
Uhhhh:
Jan 26, 2025 ~2:45 PM
IGO • A20N ROUTE AJL → DEL ALTITUDE 1,325 FT GROUND SPEED 137 KTS VERT. SPEED -768 FT/MIN
https://pilane.obviy.us/_astro/38e014bd.CJs63r7b_Z2lCImg.web...
Unfortuntalely that happens more often than I'd like, especially if the frame contains both the plane and a bird. Hoping to fix that in V2!
related: automated Las Vegas plane spotting https://www.youtube.com/watch?v=d5PtT7KdlKc
this person has some sort of OpenCV setup from a high-rise view. (bonus: Air Force One is currently there)
I'm kind of surprised that they would park it next to other planes like that. Especially with what appears to be no ground troops surrounding the plane.
>I'm kind of surprised that they would park it next to other planes like that
The extreme zoom of the camera makes everything look close together. In reality it's probably quite far from most other planes.
>Especially with what appears to be no ground troops surrounding the plane.
Why would you have them stand outside to guard the plane when the airport is already fenced off, and has surveillance/security? Even if they do need people guarding it, it doesn't make sense to stand outside. Waiting inside or in a nearby SUV would be much more comfortable for the guards.
Guards sitting comfy in an SUV is not a really good guard detail.
Surveillance cameras placed by the airport provides much better situational awareness than a bunch of bored guards standing around.
That place is crawling with cops from the gutter to the street lights.
Trump hooking up with Stormy Daniels....?
You can see from the photos how people might mistake planes for drones or UAP's.
Every single one of those pictures, even the blurry ones, even the ones taken at night, all look like airplanes.
People aren't getting airplanes confused with drones because they look the same. People are getting worked up about drones because of collective paranoia, and then they are seeing what they want to see.
I don't even know what they're afraid of, these drones aren't carrying an RPG round like in Ukraine. I guess it's not impossible.
Your last sentence negated the first and only furthered my point. Moving lights in low visibility are open to a variety of interpretations.
Um, no, I can't actually. Like, not even close to being able to relate to what you're suggesting. These are all single frames and frozen in time. Watching any of these actually moving would even reduce that possibility even further.
Not criticizing you but it will never not be funny to me that crazy people have tried to legitimize seeing alien spaceships by renaming them from UFOs to UAPs. It’s like the conspiracy theorist version of unhoused.
The problem is you need a term for "flying thing we saw but weren't able to identify and/or categorize" (which is certainly a legitimate need) but every time you start using it this will increasingly change meaning to "aliens and such are zooming around and <some power> doesn't want you to know" (completely regardless if it's factually true or not - nobody is going to want to use an alternative term someone sets aside for "crackpots" in either case) until it gets to the point people don't even associate the term with what it was created for again.
In fairness, I get what you’re saying, but I think there is some legitimacy to it.
First, I believe that this term has been introduced by the us government, so it’s not the crackpots laundering the conspiracy themselves by using a new term.
Second, I believe the reason UAP was introduced is to describe properly recorded and credibly witnessed and described phenomena (tic-tac, etc) that are not explained by any publically known craft, engineering, or science, but aren’t likely aliens either; and that is well described by UAP which doesn’t directly imply aliens that ufo does.
Now spot UFOs/UAPs?
