Anyone able to use cameras intelligently with Vera? In my setup, there’s a camera on the inside of the main door. When I detect the door opening (I’ve set the detection area to be around the top of the door/frame) I’d like to snap stills for a certain amount of time - or basically as long as there’s movement in the frame. It could also be video - I don’t really care about just that. No matter what I’ve tried I don’t seem to get the desired effect.
Cheers,
Lars
Some cameras, as well as all surveillance software, use motion activated recording. Hikvision probably has the most sophisticated in-camera software.
Face Detection. I am also interrested about knowing who is on the door. My first idea was to give RFID TAG to each family member and via a RFID reader vera could recognise who is coming in. My daughter but on idea to have a camera + some face detegtion software to do the job.
I did investicate a litle and was suprise to find out that at least google has some cloud solution for this https://cloud.google.com/vision/?gclid=CLX40Mm0z8sCFebbcgodcr4H2g. It says though “Facial Recognition is not supported”. 
Has someone done anything arround Face Detection on VERA?
The netatmo camera advertises face detection. I haven’t tried it, but saw a demo at CES of the camera identifying people vs pets, cars and other motion, and it seemed to work pretty well. They say they have an open api, so it should work with Vera, but haven’t seen anyone try to integrate it. I’d like this as well. I wonder though if in practice it will work consistently enough to be useful–seems like you’d have to have a person be looking directly at the camera?
Microsoft is having Project Oxford for this
I have noodled in OpenCV sometime back and it has lots of built in libraries for different imaging needs. They have a facial recognition library[url=http://docs.opencv.org/2.4/modules/contrib/doc/facerec/facerec_tutorial.html]http://docs.opencv.org/2.4/modules/contrib/doc/facerec/facerec_tutorial.html[/url] that I have played with and works fairly well. I had a vision of cameras being the ultimate sensor and I still feel that way. I was trying to see if I could have smart cameras that could identify people versus dogs, etc. Learn what objects should be around my house so anomalies could be identified. Given the amount of video processing and HP this was going to require I put it on the back burner for now. That said, I think a very focused effort towards facial recognition could work without being a huge effort. You will need a good up close and personal shot of their face so the camera’s placement would be important. If you notice, all the facial recognition samples you see are very close pictures. So one of the biggest challenges I ran into was getting that perfect frame for comparison. It is easy to take a picture and run it through an engine but getting that shot from a stream in an automated fashion is tough. The second thing that is needed is some kind of video processing server that will have the OpenCV libraries installed. This machine will need a good video card as well, preferably one with a lot of cuda cores. I use the cuda build of opencv and have a video card with over 800 cuda cores that I use for image processing. Trust me, you will want your video card to do the image processing instead of your computers processor. It would be cool to walk up to my front porch and have the camera identify me and unlock the door I guess. More importantly I think it would be nice if it identified when someone was at my front porch when I was away. Let me noodle a little on this and see if I can come up with something relatively simple to leverage the opencv facial recognition libraries.
Best Home Automation shopping experience. Shop at Ezlo!
© 2024 Ezlo Innovation, All Rights Reserved. Terms of Use | Privacy Policy | Forum Rules