Open Nexulo opened 4 years ago
I did that in https://github.com/robmarkcole/coral-pi-rest-server It is not necessary for my use cases. Keeping this issue open as I am sure it will be asked again
Was guided here by the coral-pi-rest-server (deprecated) - have you dropped coral support entirely?
I might add it in a branch, this whole repo is for experimentation
Ahah, was redirect here from deprecated repository too, when looking for face recognition with Coral USB ;) Hope you can do it soon ! :D
I am thinking about processing video feeds again, so might consider this
@robmarkcole, I see this new Deepstack deposit on .github just appeared 2 days ago
https://github.com/johnolafenwa/DeepStack
Any chance we will get a docker container that supports our Coral stick that can be called from home assistant without having to wedge the rest server you wrote onto Linux?
I did this port of your server and it works, but it's not optimal... Docker would be way better.
https://community.home-assistant.io/t/image-object-detection-w-google-coral-on-intel-nuc/208248
@jazzmonger the issue is that Coral, which uses TensorFlow-lite, is incompatible with the Pytorch approach we have taken in the rest of Deepstack. Feel free to continue this discussion on the Deepstack issue https://github.com/johnolafenwa/DeepStack/issues/36
Any chance this will work on Intel NCS2?
@fgonza2 similar to my reply re deepstack, I believe NCS2 uses intel specific modes, not tflite
I've unarchived the coral repo, so people can make PR to add functionality
Is the USB stick not supported (Why not?)? Will the (optional?) support come in the future?