My wife bought this zeromouse2 device (https://zeromouse.com/) which will help stop the cats bringing in prey, sorry presents.. Not the cheapest, and from what I can see (from job postings), it uses esp32 internally and connects to AWS. I havent taken it apart yet as its still in trial phase, but wondering if anyone else has looked at this device?
Sure , its expensive, but I'll be honest Im so time poor at the moment I wouldnt have the time to build one with a pi, camera, make it work, test it , train the model and so on....
At the moment I dont, its on the 30day trial period but Id like to receive the messages into HA to do clever shit (cat came in , check water bowl etc)
Also who knows, one day the company will go pop so I want a backup.
I was going to build the solution myself, I have a bunch of esp32s, a pi, I dont have a camera or prox sensor but I am a coder, access to AI skills, etc etc but I am so time poor for private projects now...
Anyway it was more curious, the solution looks good , it was a bit pricy but I am paying for time, and they are hosting the app+ai models etc so Im ok with that
Agree, there are plenty of projects online, github etc..
Software wise, I dont see an issue with the software technically but yeah time sink, also kudging it together in a way that works in the catflap, weather proof etc takes time
If you program the algo to detect face of 5 cats in the family during night without passing them mouse in the mouth and also blocking all other cats in neighborhood, I'll build the hardware
He also uses a raspberry pi, which makes the models easier to use..
The main issue (for me) isnt the coding, its been done before, and there are many projects out there. but the sample data. Labeled data with examples of prey/no prey. And if you use Codex/Claude things get easier.. For me its the hardware, sure Ive built esp32 devices, and could get them working with a camera, small enough etc, but all that will take time, time i dont have.,
ZeroMouse have done a nice job of putting that into a esp32 (instead of a pi - thats challenge) built the infra, push notifications etc, all doable just takes time to get right.. and then back to the labeled training data,.......