Next step, labelling and Tensor, then I can see if this project is going to work (at least, via webcams), and *possibly* take some steps to for auto labelling
It's working!
Even with the most basic, horribly unoptimised tensor network with a couple of minutes of training, I've work a working solution.
I've got auto labelling setup, so more training data will appear over time.
The code could use *lots* of optimisation, I'm using a jpeg as an intermediary, but I'm really pleased with the results
@nickjvturner so basically my Sure connected cat flap broke at 3 years 2 months on a 3 year warranty, and they thought it was reasonable to tell me to throw it away rather than replace a couple catches.
If the service had been exemplary for the life of the product, maybe.. But they want through several campaigns locking out attempts at homebrew local integration and cloud integration.
So the aim is to build an open cat flap you can 3d print. Because screw them.
@wishy we should talk!
I like this idea.
I have just completed a battery elimination project for the SurePetcare ‘Pet Door’
@wishy yeah, the pet door is connected, but I have to use an external locking mechanism because of the api slowness.
I was also thwarted by the pethublocal ‘fix’ they released.
I would like to use the integrated locking mechanism, but 8 round trip seconds is too long.
@nickjvturner the Api being slow is a big factor in wanting something local. That and repairability.
The bill of materials works out about the same as a new catflap. If they were just a little more open to the home automation community, that could be money in their pocket. But alas...