Beyond Security: AI Cameras and the Smart Home That Can See

Smart Home Devices
120 followers
9 Video Views·Jun 11, 2026
I installed the Reolink OMVI 3i PoE camera for a simple garden timelapse… and ended up exploring something much bigger: what happens when CCTV goes beyond being a recording device and becomes a visual sensor for the smart home?
AI Generated·May contain errors
One of the most futuristic questions I can now ask my smart home is,
have I left my cup outside?
That sounds tiny, almost silly,
but I think it points to a much bigger shift in what cameras are about.
Because for years, home cameras have mostly been treated as security devices.
Something moves, they record it.
Something happens, you go back and check the footage.
Now, if you've seen any of my other videos,
you'll know I use my cameras for all sorts of non-security reasons.
But this time, I'm going even further.
Because now that AI can look at those camera views
and actually understand what is in them,
the camera is no longer just a pair of eyes observing a scene.
It's the brain as well.
For six hours and 10 minutes.
Congratulations, you're quite inept.
The lawn remains untidy.
The workshop looks a wreck.
The probability of finding three unfinished projects has risen again.
Now, this all started with me wanting to install my new Reolink camera
to take a simple time-lapse of myself doing up the garden.
Well, garden is a pretty generous word.
Rubbish dump is more accurate.
But it quickly became something much cooler.
My new camera is the OMVI 3i POE.
It's got both the wide-angle camera and a 360-degree camera.
So although I was only trying to capture my yucky garden,
its range of vision captures much more than that.
Normal life happening, like bikes, deliveries, weather, pets, objects being left around.
And when I combine this view with my other Reolink cameras around the house,
the picture just gets even richer.
Because they can also see my washing line, the fruit trees, my workshop,
the children's play area, and all sorts of other details around the house.
So there I was, looking forward to a bit of gardening and getting the camera set up,
when I happened to spot a message from Pedro in my Discord channel.
He was super excited about what he'd been doing with the Reolink cameras that he'd bought.
And in short, he wasn't just recognizing objects.
He was turning objects into a series of sensors
that he could feed into platforms like Home Assistant.
So he wasn't just recording his devices running.
He was turning their lights, their screens, their states into data that he could read.
I thought this was brilliant.
And I was itching to see how far I could take this.
Would it be possible to use this technology to manage me
and prevent me from doing stupid things like leaving the washing out in the rain
or mowing the lawn in the shape of a football?
Would it help us look after our things better by
telling me that I've left the hammock out before high winds
or noticing that I've left a power tool on the lawn?
I had so many ideas I wanted to test.
So after installing my new cam and mulling over some ideas while doing a bit of gardening,
I loaded a snapshot image from the CCTV into several AIs
to see how much information they could derive from a camera shot.
And the result was astonishing.
It picked out every object in the scene.
And it didn't just give me generic descriptions,
but the brand, the product type,
and all sorts of information I hadn't even spotted myself.
It noticed the weather.
It spotted unusual objects in the scene.
Counts of objects.
What the objects were made of.
Their colours.
Descriptions of plants.
And not just vague descriptions.
I was able to ask things like,
how long is the grass?
Or does it look like the grass needs mowing?
Some of the descriptions were moving into Dexter territory.
Possible wheel slash drag marks pattern.
That's definitely not from any nefarious behaviours on my part.
After finding out that the AI can turn an image into useful data,
what followed was two days of aggressive nerding
to produce a multi-layered system
that acts as a kind of bouncer slash butler
slash accident prevention combo for my home.
It has almost god-like powers
to not just record the past,
but anticipate the future.
Which is why I called it Janus.
Now I'm not one to reinvent the wheel.
So the first layer of AI is already happening inside the Reolink world.
The camera is not just sending me a dumb video feed.
It can recognise things like people, vehicles and animals.
And depending on how you configure it,
those detections can become alerts or events
without me building anything custom at all.
You can also create zones to monitor for things moving
or detect if an object has left an area
or check if an object has arrived and all sorts.
So even at this first layer,
the camera is already doing more than recording.
It is starting to label parts of the world.
And this is where you do need to select a good quality camera
so that it can pick out all of the details.
This one has 18 megapixels of eyeballage across three cameras.
But let's move on to the second layer.
This is where I combine the data coming from the camera
with all the other state data from sensors in the house.
Things like the doors, weather, lights, occupancy,
automations and notifications.
You won't get to this point
if the brand is trapping you in their ecosystem.
But Reolink allows you to connect to Frigate
and then into Home Assistant.
But to take the dial to 11,
I built an experimental third layer.
Janus periodically takes still frames from the cameras,
stores them as a kind of visual memory
and then asks an AI vision model
to monitor specific watch items.
The reason I want to show it
is not because everyone should copy this exact system,
it's because it demonstrates the bigger idea.
Once the camera can provide open access,
the smart home can start asking questions
about the visual state of the world.
The kind of questions which inspired
this experimental detour in the first place.
Like, is there washing on the line?
Are there objects on the lawn before the robot lawnmower runs?
Is there a parcel by the door?
Are the bins still in the bin store?
Is there anything unusual in this scene?
Those answers can then become normal smart home state.
Not a video clip that I have to watch,
but a piece of information that the house can use.
And this is where the camera goes
way beyond being a security vice.
Way beyond even being a sense,
but a sense with a brain.
A camera with an AI layer
can answer messier and vague human questions.
What is on the table?
Does this area look clear?
What has changed since this morning?
Is something out of place?
Not only that, but because Janus
shares everything as sensors with Home Assistant,
you can make infinite automations.
So if my weather sensor detects wind above a certain speed
and the hammock is out, alert me.
I don't need to go through all the horror stories
of trampolines and parasols and the rest
to show why this could be life-saving.
I made a whole video about my bin reminder,
but what about reminding me to bring them back in
so I don't block the pavement for everyone?
This can see which ones are there
so that it can remind me to collect them.
That's why things like resolution,
reliability and integration are important.
High resolution means that there is more detail
for an AI model to interpret.
Pan, tilt and zoom mean that you're
not locked into one static view forever.
Power over ethernet, which is what my new one is
and is the type that I normally go for,
means that the camera can become
part of the home's infrastructure
rather than just another battery that needs attention.
And the really key thing is openness.
And by that, I mean support for systems
like Home Assistant, Frigate
and other local tools like Janus.
Open protocols mean that the camera can feed
whatever system makes sense now
and whatever system makes sense later,
which means the camera can be used
as part of a wider smart home architecture
rather than being trapped inside one app.
I started this video by saying
that I wanted to do up my garden,
but as you can see, I got a little bit sidetracked.
But the opportunities truly are endless.
Everything is normal
Everything except for you
Janus is observing
Your chaotic point of view
The gate was left open
The washing survived the rain
No action is required
You're clearly still insane
You misplaced a nail gun
I located it like that
You searched for seven minutes
I searched for 0.3 seconds
This is not a competition
Although if it were, the outcome would be
Statistically predictable
Everything is normal
Everything except for you
The parcel has been delivered
The cats are both inside
The mower encountered difficulty
I told you it would again
A report is generated
You're clearly still insane
The cup beside the workshop
Has been there for four days
The same applies to the cable reel
And the paintbrush
And the ruler
And the seventeen objects
You described as temporary
Processing
Observing
Categorizing
Optimizing
Admiring
Judging
No, not judging
Observing
Final
Everything is normal
Everything but you
The world is full of data
But I still can't grasp you
The birds have left the feeder
The sun has crossed the sky
You have been extremely interesting
To statistically quantify
Thank you for your cooperation
Thank you for once again
Continue being abnormal
You're still somewhat insane
As an aside, I will post some of my gardening progress for Patreons
and I'll share some of Janus's analysis of my progress too
But one final thought
For me, this has started to move my thinking of where the smart tech should be
For example, my lawnmower has all this tech built into it
Bump sensors, ultrasonic sensors, cameras and things like that
And these are to help it detect items on the lawn
But I can do that myself
And much more
It has sensors to detect if it's raining
Well, Janus can detect if it looks like it's going to rain
In other words, see the future, not just the present
Even my 3D printers and tools
The AI can see what colour filaments are in the machine
It can see if I've left the AMS open
Instead of installing smart contact sensors on every single window
Well, the AI can see if the windows are open
Now it can start to understand what it's looking at
And that changes what a camera is for
The smart home has had sensors for years
What it's getting now is sight
And the potential is pretty endless
But what ideas do you have?
I did leave my cup outside
We'll vow
Everything is normal
Everything but you
The world is full of data
But I still can't grasp you
The birds have left the feeder
The sun has crossed the sky
You have been extremely interesting
To statistically quantify
Thank you for your cooperation
Thank you for once again
Continue being abnormal
You're still somewhat insane
