Hello,
I am still trying to get an edge impulse model working on my pc. Is it possible to convert the files to a .pb so I can run it in a python script?
I know edge impulse was created for edge devices (as the name suggests lol) but, it would be really great to run in on my local pc.
Hi @kevin192291 you can get the SavedModel file from Dashboard in the downloads section.
Note that that is only the neural network. If you’re on Linux you can use https://docs.edgeimpulse.com/docs/edge-impulse-for-linux which also has all DSP code and has a Python SDK!
Hello!
Thanks for the tip! It works great! The trouble I was having before was that I didn’t update the NPM package!!! I have 2 quick questions though, the first is how to get a video stream from my IP cam? Is this possible? Here is my attempt:
# input video
camera = cv2.VideoCapture("rtsp://root:toor*@192.168.2.179/onvif-media/media.amp")
# camera = cv2.VideoCapture(videoCaptureDeviceId)
# backendName = camera.getBackendName()
# w = camera.get(3)
# h = camera.get(4)
# camera.release()
next_frame = 0 # limit to ~10 fps here
for res, img in runner.classifier(camera.read()):
...more stuff...
It seems that runner.classifier() only takes an int, and that is the number of webcam directly connected to my pc, is that correct? Is it possible to grab other stuff than just webcam images?
and the 2nd question is about hardware acceleration, When I downloaded the model, it says it is not enabled. Is that something that would change in the future? Or is it even needed? (sorry, I don’t know too much about this stuff, but I am trying!)
Thank you very much!
-Kevin
Hi @kevin192291, see the code here: https://github.com/edgeimpulse/linux-sdk-python/blob/07704b32792c630c64dd315d05d93a6306aae50d/examples/image/classify-image.py#L49 - this should work when you have an image (as you have in the above situation when calling camera.read() I think).
and the 2nd question is about hardware acceleration, When I downloaded the model, it says it is not enabled. Is that something that would change in the future? Or is it even needed? (sorry, I don’t know too much about this stuff, but I am trying!)
Where are you running this and can you post the exact message you see? All Linux systems should have hardware acceleration enabled these days.