IAmATeaf
Known around here
Will Linux or freebsd lower power usage of a PC?
You should join the homelab reddit group - there's lots more linux users there.Thanks I will keep this in mind for future use and now I know what my professor was alluding to in class.
Thanks for sharing! I am just getting started in my setup. I am half tempted in grabbing one of those AI google coral usb devices to take off processing load from my system. Are you processing everything locally or are these cloud based AI solutions?If you want something with powerful AI then definitely take a look at Frigate. When used with DoubleTake to train the facial recognition it's great. DeepStack is not impressive. CompreFace is amazing! My system will text me with something like "Jenny the cat 98% pooped on the south lawn at 2:32am. All open source.
My stack looks like:
Home Assistant - UI
Frigate - DRV
DoubleTake - Train/personalize AI
CompreFace - facial recognition
Check out the homelab and home automation groups on reddit - this is a good place but definitely youll find tons of information there.
Yeah I see a few sellers on amazon for just over $100 - I am in the US btw.I'm a happily paranoid. I do everything locally. My video is saved on a ZFS NAS (TrueNAS). I have a 2nd cheap NAS (OpenMediaVaulton on an Odroid H2) at a remote location that has backups of high probability events. Just in case someone brakes into my house and takes my main system. That Google Coral AI ASIC is literally x100 faster then my Xeon CPU! Can put it on a Raspberry Pi. Unfortunately they are hard to get at the moment. Normally they should cost about $70.
Does a Google Coral in any way shape or form require a connection to Google? If so I already don't trust it.Google Coral
no.Does a Google Coral in any way shape or form require a connection to Google? If so I already don't trust it.
No way. If they did something like that nobody would use them. I have spent hours in WireShark looking for calls. Last people I want to have my data -from inside my house- is Google.Does a Google Coral in any way shape or form require a connection to Google? If so I already don't trust it.
Any chance you could post a quick tutorial on your setup?If you want something with powerful AI then definitely take a look at Frigate. When used with DoubleTake to train the facial recognition it's great. DeepStack is not impressive. CompreFace is amazing! My system will text me with something like "Jenny the cat 98% pooped on the south lawn at 2:32am. All open source.
My stack looks like:
Home Assistant - UI
Frigate - DRV
DoubleTake - Train/personalize AI
CompreFace - facial recognition
It's very much like this one. The Asus card used here is also using the Coral chips. Its WAY overkill though. You only need one chip, not a card full of them. I have trained my system to turn on the sprinklers when certain animals walk on my grass. Dogs, cats or anything that might take a dump. If somebody not in my family goes close to my car, a light starts flashing and I get a text msg.Any chance you could post a quick tutorial on your setup?
You can also check out this guys channel if you are more into using blue iris and a windows environment. This solution uses your local hardware for AI processing instead of a coral.Any chance you could post a quick tutorial on your setup?
This is with Blue Iris that the OP is looking for alternatives on....and these videos are so old that the AI has since been integrated with BI and is way easier to implement now than when these videos were made and it was considered more 3rd party tools...You can also check out this guys channel if you are more into using blue iris and a windows environment. He is using a cloud based solution for AI processing tho instead of a coral.
Here is one where he does AI locally but on his own hardware, again in a windows environment:
I didn't reply to the OP, I replied to someone else who is just getting started. Interesting you mention that AI is integrated with BI but I do not see that as a feature on BI's web page, you'd think that'd be all over since its a hot buzz word. I am just getting started and can be totally wrong.
can you show a youtube video for reference demonstrating the new integration in BI?
Blue iris is developed by one person. He does not update the website often. This is the new AI integration for BI, previously it was with deepstack.I didn't reply to the OP, I replied to someone else who is just getting started. Interesting you mention that AI is integrated with BI but I do not see that as a feature on BI's web page, you'd think that'd be all over since its a hot buzz word. I am just getting started and can be totally wrong.
A bit OT but I after watching the video I saw something called "deepquest" in his settings. Some googling lead me back to a 2019 thread here by gentlepumpkin. My guess is Blue Iris uses a online API to access most likely Google Tensor Processing Units (Edge chips in Coral available 2018) for a monthly fee. If so this would be very effective at identifying a huge list of objects - toilets to cows.- with very few false positives. Not faces to names though. Best of all you don't have to try and buy the Google Coral chips. As for identifying "Gary the mailman" it seems they are using DeepStack. I don't see how they are teaching the AI faces to names. DeepStack is... mehh IMHO. CompreFace is a lot better. When "Gary the mailmain" is seen on my cameras I just drag and drop a few pictures of him into DoubleTake and it teaches the AI who he is. Same with the evil cat "Jenny" down the road. I don't want my sprinklers blasting my kids cat. Just Jenny the pooper cat. Here is the ipcamtalk thread I found from 2019 btw.can you show a youtube video for reference demonstrating the new integration in BI?
That is one of the many 3rd party tools before AI became integrated with Blue Iris. Some people still run that for the finer granular aspects, but most are going to the integration with BI. First integration was DeepStack but it has since switched to SenseAI/CodeProject.A bit OT but I after watching the video I saw something called "deepquest" in his settings. Some googling lead me back to a 2019 thread here by gentlepumpkin. My guess is Blue Iris uses a online API to access most likely Google Tensor Processing Units (Edge chips in Coral available 2018) for a monthly fee. If so this would be very effective at identifying a huge list of objects - toilets to cows.- with very few false positives. Not faces to names though. Best of all you don't have to try and buy the Google Coral chips. As for identifying "Gary the mailman" it seems they are using DeepStack. I don't see how they are teaching the AI faces to names. DeepStack is... mehh IMHO. CompreFace is a lot better. When "Gary the mailmain" is seen on my cameras I just drag and drop a few pictures of him into DoubleTake and it teaches the AI who he is. Same with the evil cat "Jenny" down the road. I don't want my sprinklers blasting my kids cat. Just Jenny the pooper cat. Here is the ipcamtalk thread I found from 2019 btw.
GitHub - gentlepumpkin/bi-aidetection: Alarm system for Blue Iris based on Artificial Intellience.
Alarm system for Blue Iris based on Artificial Intellience. - gentlepumpkin/bi-aidetectiongithub.com