Jump to content

The easier solution is to get motion detectors (as the video mentioned) and then just hire a journalism major (JM) to walk around and flap his arms to keep the sensors active.

This serves several functions:

 

(1) Keeps riff raff (like journalism majors) off the streets

(2) Easier to implement

(3) no worrying about e-waste cluttering up the landfill. When you wear out one JM, throw them away and get another.

Piece of cake 😉

 

(reference: https://dilbert.com/strip/1995-02-22)

 

 

NOTE: I no longer frequent this site. If you really need help, PM/DM me and my e.mail will alert me. 

Link to post
Share on other sites

Yeah.... I don't know, I wouldn't be comfortable knowing that every inch of my house is being watched by cameras. Even if nothing is logged the cameras are there and the images can be accessed. That basically means there is no privacy from Linus in the house. (He would be able to get access to the camera feed any time he wants to) while his kids might not mind that now, I am not sure what they'll think about that once they turn teenagers...

I get why this feels so cool and exciting, but this can be abused so badly. This is one of the things that I feel are the wrong way to go with technology. I would've taken the cellphone route simply because this would make it possible to opt-out of it anytime I don't want to be tracked.

Link to post
Share on other sites

3 minutes ago, XWAUForceflow said:

Yeah.... I don't know, I wouldn't be comfortable knowing that every inch of my house is being watched by cameras. Even if nothing is logged the cameras are there and the images can be accessed. That basically means there is no privacy from Linus in the house. (He would be able to get access to the camera feed any time he wants to) while his kids might not mind that now, I am not sure what they'll think about that once they turn teenagers...

I get why this feels so cool and exciting, but this can be abused so badly. This is one of the things that I feel are the wrong way to go with technology. I would've taken the cellphone route simply because this would make it possible to opt-out of it anytime I don't want to be tracked.

Yeah me neither. 

https://www.lttstore.com/

1990 M3s are the best looking things ever made.    

^This statement has been retracted^
2020/2021 BMW S1000RRs/Ninja H2s are the best looking things ever made. 

Don't ask to ask. 
If you want me to see the reply, @XGoodGuyFitz(aka me) and/or quote me.
Thanks!

 

Link to post
Share on other sites

As someone who works in building-automation sector this is very interesting. Presence detection such as "how many people are in a room" is not easy (=cheap). For just presence detection there are sensor available that detect people that are sleeping or sitting at a desk working.

Some fairly good sensors for these purposes are made by a company called "Steinel". Their products such's as "HPD2" and "True presence".

The "True Presence" detects people even if they are still.

The "HPD2" is a camera that locally analyzes picture and gives out information if it detects people. It can also give out the number of people detected.

So I urge you to check those out. They are not cheap by any means!

Link to post
Share on other sites

I wrote some face detection software a decade ago with opencv for my bachelors degree. Had to do some trickery, to be able to run in real time on a dualcore.

 

From my experience from back then:

  • The camera has a huge impact, not because of picture quality, but because of speed. Most webcams create a framebuffer in the camera beforce starting sending pictures to the computer. This creates lag. I settled with the playstation eyeToy because it was fast. Be careful in your choice of camera.
  • You dont need a high resolution for object detection. 320x240 was enough for face detection for me. Face recognition might need some more. This saves a LOT of processing power AND increase privacy a little.
  • Greyscale pictures are enough for object detection in most cases.

The goal here is to save energy and create convenience. Even if it is "only" 52W for the card, I would try to reduce the consumption with a lower picture resolution and less FPS. It is 52W 24/7. I would recommend testing it with greyscale images and what resolution you actually need for your cameras to work, just to save money and energy.

Link to post
Share on other sites

Think this project might be doable with a single $40 M.2 TPU.

 

For HVAC efficiency, a much lower polling rate would be fine - perhaps poll every minute, but turn off HVAC if no person is detected in the last 15 mins. Further, person detection might work acceptably with much lower resolution snapshots possibly as low as 100x100

Link to post
Share on other sites

34 minutes ago, GodAtum said:

I'd like top do this in my home to use facial recognition to disable my alarm as I'm always forgetting! Any detailed instructions?

Easier solution:

 

Set your alarm to the loudest volume it'll do.

Tune the alarm to the musical genre you hate the most.

 

I guarantee you will never forget to turn the alarm off again.

NOTE: I no longer frequent this site. If you really need help, PM/DM me and my e.mail will alert me. 

Link to post
Share on other sites

A lot of AI workloads can be made far more efficiently with regular code. There is however certain things that an neural network is the easier solution compared to making code. (Generally anything that is sufficiently complex for one to not logically figuring out any suitable way to approach the problem logically. This simple "check" is however not indicating that a neural network is suitable for the application, the programmer/development-team can just be dumb/unskilled and not know how to approach the problem logically to start with.)

 

However doing detection of people in a room is an interesting application.

 

And to be fair here, I don't think a camera is needed. Motion sensors goes a long way, if you detect motion in a room, and it disappears without causing motion in any other room, then the person is still in the room, or they climbed out the window like a freak. This however is less applicable if there is multiple people in the house, and counting people is a bit harder especially if one's motion sensors aren't indicating direction. (Though, I have worked with IR motion sensors that trigger if one sits still a few meters away and just breaths, or slowly shift one's position. A person sleeping isn't a problem here. So general presence isn't hard to detect.)

 

Going the full AI route seems a bit like hammering in a screw, or in this case, being excessively lazy and not thinking about how to implement something with a few hundred lines of code. (Though, there is other advantages with image detection and tracking of people, but that isn't really useful for general HVAC applications.)

Link to post
Share on other sites

Damn i should have gone for the ROG STRIX 3090 WHITE instead because damn that card looks better than the SUPRIM X 3080 TI i have..

white.PNG.0287538c1978aca17025fa8ebda63c3a.PNG

Useful threads: PSU Tier List | Motherboard Tier List | Graphics Card Cooling Tier List ❤️

Baby:  ASUS TUF Gaming B650-PLUS WIFI | AMD Ryzen 9 9800X3D | Thermalright Peerless Assasin 120 | ASUS TUF Gaming GeForce RTX™ 4090 OC Edition 24GB |  G.Skill Flare X5 DDR5 6000mhz 32GB  | Corsair RM1000xKingston FURY Renegade 2TB | WD_BLACK SN750 | Samsung EVO 850 | Kingston A400 |  PNY CS900 | Lian Li O11 Dynamic White | Display(s): Samsung Oddesy G7, ASUS TUF GAMING VG27AQZ 27" & MSI G274F

Link to post
Share on other sites

it scary to see you with out a beard.

where going to leave this video at that...

MSI x399 sli plus  | AMD threardripper 2990wx all core 3ghz lock |Thermaltake flow ring 360 | EVGA 2080, Zotac 2080 |Gskill Ripjaws 128GB 3200 MHz | Corsair RM1200i |200tb raw | Asus tuff gaming mid tower| 10gb NIC

Link to post
Share on other sites

re detecting people in your house.

In a venue we manage we was looking into something similar. With high bay lighting you can get a microwave sensor. These sensors detect if someone is in the room even when they are not moving. Maybe it can be adapted for your home application.

Link to post
Share on other sites

Hey everyone, I'm David, the creator of Double Take, one of the open source projects featured in this video. It was an honor to see Linus use my project and I'm happy to answer any questions if anyone has issues getting Frigate or Double Take to run. You can also message me on Discord if that's easier too!

Link to post
Share on other sites

12 hours ago, Hobox said:

How much does something like this cost? Did I miss a mention in the video? Didn't see any rough numbers when googling

As demoed in the video, MANY thousands of dollars. The featured AI accelerator alone is over $1k.

 

Granted, if engineered to a more reasonable price point, you could use the $25 M.2 version in an unpopulated M.2 A+E slot of a spare PC, and have similar results at a fraction of the cost and power consumption.

Link to post
Share on other sites

12 hours ago, Tumeksi said:

As someone who works in building-automation sector this is very interesting. Presence detection such as "how many people are in a room" is not easy (=cheap). For just presence detection there are sensor available that detect people that are sleeping or sitting at a desk working.

Some fairly good sensors for these purposes are made by a company called "Steinel". Their products such's as "HPD2" and "True presence".

The "True Presence" detects people even if they are still.

The "HPD2" is a camera that locally analyzes picture and gives out information if it detects people. It can also give out the number of people detected.

So I urge you to check those out. They are not cheap by any means!

Another more affordable home solution is using a thermal camera and ESP32 board to detect a heat signature. Much cheaper, more power efficient, and less invasive, but wouldn't be able to detect specific people, and could potentially have issues in environments with existing variable heat sources.

Link to post
Share on other sites

12 hours ago, Texas_Ace said:

Anyone got links to all the software he is using? I am very interested in this type of real world, locally hosted, open source, application for AI but the video glosses over a lot of the particular software in use.

Home Assistant

Frigate

Double Take

Coral

 

Link to post
Share on other sites

17 hours ago, daddysmackdown said:

What cameras are being used? 

14 hours ago, jhoglund218 said:

I am also curious about this. Last I remember he was running ubiquity cameras, but I’m not sure if that’s still the case. 

I run all the programs mentioned in this video with Eufy, Wyze v2, and Reolink cameras. As long as your camera can output a RTSP stream, you should be fine.

 

18 hours ago, Hobox said:

How much does something like this cost? Did I miss a mention in the video? Didn't see any rough numbers when googling

I run all the programs mentioned on an Intel NUC8 with 16GB of RAM and I use two Google Coral USB Accelerators with Frigate which take a lot of the load off the NUC. You can definitely run all this on a sub $1,000 machine.

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×