Edge computing Local image Analysis gateway

It would be cool if you allowed a feature that could send images to a local and More powerful edge device for image a analysis.
That way my “stuff”. would not have to go to “the cloud” or another foreign server to do image processing / advanced detection. Also reaction time would be faster. I would pay 50-75 bucks to have an advanced analysis gateway on Prem that could handle if you offered it. If not, and you opened the cams up locally via say rtsp or whatever, I could use raspberry pi and google tensor flow toolkit to have advance image analysis at the edge for any images where motion was detected by the camera. I could by exception, more deeply analyze each acquired image and do all Kind of neat things.

Don’t forget to vote for your own #wishlist topic by clicking the blue “vote” button!

You know, when I posted this, I had not looked at all the other awesome ideas posted by others. After reviewing many wish list items I thought I would add to my own post above.

So the image analysis edge compute node I proposed would be a more capable platform with headroom for additional capability / customization / scripting. There are many easy languages out there, Go, FloGo, TinyGo, Scratch, Lua, and even Python, … And… i think for 50-75 bucks Wyze does not have to pinch pennies and get an under powered device. Heck, if it does the job of letting me control my destiny for a one time fee i might even be okay with paying more. One time!

Also, It seems to me in reading sooo many great posts that Wyze, for whatever reason, feels uncomfortable creating an OPEN API on its underpowered devices like the cams, sensors, bulbs, etc. Fine, why not expose the API at this proposed edge compute module. Then you can keep the last mile proprietary for whatever reason you have,. I did buy 4 cams, a bunch the sensor kit and a bunch of sensors, thinking I would create me own robust security systems, Gotta tell you, relying on subscription services, shipping my data else where and have a very limited set of integration options or ecosystem is, to put it mildly, a real bummer. And IFTT is a nice idea, but slow

Here is short time to market idea… if developing the compute module is too hard. Just sell a branded raspberry pi or new arduino nano BLE with your integration SDK, and exposes API’s to cam, etc which you charge a premium for. That’s another idea I would also pay up for. Wyze avoids the hardware development risk altogether! With this PI or ??? device SDK people could add whatever the heck they want, MQTT, blah, blah…

Having this premium integration option would allow people like me that like to tinker, to spend premium one time coin, for an onsite integration / compute node with API command and control capabilities. For others who like the cloud stuff because its simple, great! Best of both worlds.

Oh, and one last thing did I mention I hate subscriptions. :slight_smile:

2 Likes

Kenneth, I’m right there with you, and just before I saw your post I saw this interesting project:

Quoting: “The software is set up using Tensorflow on a Raspberry Pi 4 model B. A Google Coral Edge TPU was used to help accelerate the image recognition process.”

and immediately thought; Wow!, it would be cool to leverage this technology for a non-subscription option to person detection, and a cool video processing hub to include video capture to either local, or a cloud storage provider of choice that some of us are already paying for.

+1 from me!

I would like this :smiley: