Monitor your security cameras with locally processed AI

Are you using this with Home Assistant?

I'm trying to integrate this, but the HACS integration does not seem to work with my HA because the get.hacs.xyz server is misconfigured.

  wget -O - https://get.hacs.xyz | bash -
  Connecting to get.hacs.xyz ([2606:4700:20::ac43:4465]:443)
  28EBD0AA71710000:error:0A0000C6:SSL routines:tls_get_more_records:packet   length too long:ssl/record/methods/tls_common.c:662:
  28EBD0AA71710000:error:0A000139:SSL routines::record layer failure:ssl/record/.rec_layer_s3.c:687:
  ssl_client: SSL_connect
  wget: error getting response: Connection reset by peer

I am sorry to be that guy, and I think it is good that you realized it your self, but how could you trust them with your videofeeds in the first place?

Like, I remember thinking the GNU guys were hippie crackpots. But it was like 15 years ago and I have forgot how to relate to that feeling... it is like realizing all my colleagues are not using adblockers and visit sites with ads. I just can't understand.

How did you get the Tapo cameras to play nice in rtsp mode with frigate? I found that even one camera did horrible things to the wifi. Even with one camera per AP per band, they caused trouble.

Not to nitpick but you're only really guaranteed privacy unless you know there's only a wired connection. If it has wifi the camera could hop onto a nearby open network and do whatever it wanted without your knowledge, assuming evil enough firmware

Usually I would agree with you, but this is an incredibly common initialism, used by not just people in the industry, but also by consumers. Sure, it may not be as widespread as VHS (global) or API (tech-adjacent), but anyone who is in the market for this software already knows what NVR means.

Most people would know the term from either being quoted or looking up CCTV solutions, all of which, unless they're fully "cloud-based", come with a component that is called the NVR. You wouldn't even consider this if you weren't aware of the concept. If NVR means nothing to you, Network Video Recorder doesn't mean anything to you either. This is meant to be a replacement for closed and inflexible hardware boxes that are sold together with security cameras, and the name of those boxes are "NVRs".

Most stores will just market the devices as NVR or NVR Recorder (I know). If you google it, you get your answer immediately.

One good reason for cameras. They promote civil behaviour.

Since I installed a visible security camera above my front door I never had couriers throwing packages, they very rarely not show up and claim "no one was home" and so on. Also I had a neighbour damage my fence every single time he was doing his farm work (plowing, harvesting). In addition he would use an unfenced portion of my property as a turning place leaving deep/huge tire marks and did other silly shit like that despite me asking him many times not to do it. Once I installed cameras it hasn't happened once.

Then there are other practical reasons, I can review the recordings to find out which way my cat went if he is gone for a long time, or I can check is he waiting in front of the door in the middle of the night without having to get out of bed. Also my cameras resolved a mystery how one of my cats got injured once (hint - deer really don't like cats).

Finally, let's say there is a huge storm forecast and I'm away. I can check remotely everything is fine.

Finally, cameras are very good for insurance purposes. At least in my country insurers are known to weasel they way out of paying very often. If you have an actual recording that is much more difficult for them to do.

The only issue I have with most reasonably priced Cctv cameras is that they go towards more megapixels when they should go towards more IR sensitivity. Almost every consumer grade camera can be defeated at night if a subject is moving quickly. The picture will be smeared. So for ID purposes I use lower resolution more "professional " cameras.

As for open source, I've been using ZoneMinder with local (and on camera) AI for ages.

> They increase stress level and felt insecurity. They do not make you feel secure, say psychological studies. You probably think more about burglaries and dead spaces in your setup and actively monitor for these in your daily lives, where for 99.8 % of people this should be a non-topic.

Oh wow, I didn’t know I felt that way! I’m glad you were able to tell me what I feel.

You are making a lot of assumptions about why people have them.

People have different dispositions, live in different environments with different levels of support from law enforcement and face different threats. I live in a remote area and am regularly away for extended periods of time. I’ve spent years with and without any security cameras and I’m generally more content when I have a few keeping an eye on the place.

My doorbell has a camera that records locally.

When the doorbell rings I get a notification on my desktop and phone with a relevant image captured a few moments before the button was pressed.

Then I can determine if it's something I need to put my pants on for.

Mostly it's just fun and easy to add cameras around your house. Then you can do stuff like have the LLM count birds it sees or ask it "are the dogs in the back yard" etc.

Like with all home automation, you should use it to solve problems you have, not problems you want to have.

Here are some ways I use security cameras:

Check if my colleagues are in the office or not (and if they are in the middle of a live recording). Check on my plants while I'm away. Check if there is a free parking space. Check if I left something at home or in the office.

I'm not really thinking about crime, even though they are called 'security cameras'.

I've got two and will probably add a third.

The one pointed at the driveway sends an alert to my phone when someone visits. It's handy because I can't hear the house from my office so I often don't realize when we have guests over.

The one in my back yard is for security. I don't obsess over it, but if something went missing from my workshop I'd check the recordings. I'm not worried about traditional thieves, but I've got a couple unsavory family members.

Your argument is like “If we don’t do covid testing, we’ll have no covid cases”

You can integrate it with home assistant to send notification on your phone (or run any other automation) when it detects any specified objects.

I have set it to send me notification if any person is detected in my front yard, drive way or back yard after I have "armed" my alarm at night. I am thinking of also sounding am alarm on my home speakers.

Frigate, when configured properly, has a really low false positive rate. I have only seen 2-3 false positives in the past one year. And if rarely ever misses. So it's something you can rely on.

Probably a "scramble suit" [0] or just a tshirt or hoodie with patterns engineered to escape AI recognition [1]

[0] https://en.wikipedia.org/wiki/A_Scanner_Darkly [1] https://medium.com/data-science/avoiding-detection-with-adve...

Looking in their github, it says that it uses openCV and Tensorflow. The motion detection is done with openCV and will be immune against any attack unless you move so slow that you are under the detection threshold.

Tensorflow for the object detection doesn't do any OCR thus written instructions dont work. However, according to the website the system has a limited list of objects it detects. So maybe disguising yourself as a walking tree might prevent detection.

More like, wear a full body raccoon suit.

I like the idea, but no.

They have a two-stage approach, first motion detection with - I think - OpenCV and then afterwards object detection of zones of interest with different object detection models, depending on your hardware.

It supports Coral TPU, Halio Accelerator and most GPUs. I think AMD is still the worst, since ROCm is not available on iGPUs.

Afterwards, they provide/support models like edgedet (Coral), YOLO-NAS, YOLO, D-Fine or RF-DETR.

They also offer paid access to a specially trained version of YOLO-NAS where you can also train your own images.

waves hand

"These are not the detections you are looking for."

Maybe if ring or whatever major manufacturer popularly rolled this feature out and criminals could easily ID ring cameras

> no more small animals waking me up in the night.

not waking you, but it is cool to have a collection of animal photos. Sort of amazing there's a hidden world.

This is becoming a real problem because the drivers/software for the Coral AI boards is yet another example of Google Abandonware(tm) which has a hard dependency on a Paleolithic-era version of Python. Comically, the hardware is still sold.

In so many words if you expect to use the Coral boards you are stuck on EOL versions of Debian/Ubuntu - which have terribly old video drivers and missing kernel GPU support. There's a good chance your modern GPU - even well-supported Intel ones - won't work.

Imagine buying new hardware in 2025 whose software still required Windows 7.

Mines been getting worse.

Been running about 2-3 years, was mostly fine before but now I get constant false positives from the children's garden toys, scooter left in the garden, pirate flag waving etc.

I don't submit false positives for privacy reasons but I'm looking at trainingy own model. I've got years worth of positives/negatives to train on.

That "subscription" is one which I gladly pay due to multiple reasons:

1. It supports the developers(s) 2. The price can be directly attributed to cost for training 3. You can keep the models you trained during your subscription indefinately

That's pretty much the opposite to AgentDVR. I don't need hosted services for remote access or push notifications - I can do that myself. But if I want to abide the license terms, I need to purchase a monthly subscription for remote access over my own VPN.

is it possible to not just recognize people but identify them? (with registered pictures beforehand ofc)

I'd check out the OpenCV documentation and examples. This is basically what I use for face recognition in videos[0]; for recognising cars or other objects, you'd probably want to either train your own model or use something like OpenCV's YOLOv3 (example: [1] but you'd need to steal the video reading code from the first link[0])

[0] https://github.com/ageitgey/face_recognition/blob/master/exa...

[1] https://github.com/deveth0/python-opencv/tree/master/objectD...

Ask a decent (non-free) AI this question, and I bet it can make you a python script to load a video and output which timestamps show a driver holding a phone.

I also don’t know, but this might be useful: https://en.wikipedia.org/wiki/You_Only_Look_Once

I am using Motion [0] since years. At least for basic stuff, is easy to configure and very flexible. For more advanced configuration, it required a bit of tuning.

[0] https://motion-project.github.io/

Polling HN: is there any upgrade to Coral? It's 5 years old at this point, and with the explosion of AI apps & HW acceleration, I'm surprised there doesn't seem to be anything to update Coral's niche, of an IO-attached NPU.

For on-camera AI, I'm aware of OpenMV https://openmv.io/ and their recently-kickstarted N6 & AE3 https://www.kickstarter.com/projects/openmv/openmv-n6-and-ae...

a) How many LAN cameras b) How many WiFi cameras

are you using with only one Coral USB dongle at the same time (plugged in the PN50 NUC) and get successful object or person identification with frigate? And why telegram? Is it connected to frigate only for notifications resulting from the identifications?