Raspberry Pi Image - First Run

ClintCrowe

Hi all, I have the image installed on a RP4. The directions are a little dated and I dont find the demo file demonstrated in there. I see an autostart. How does that autostart? What else can I be running with just the base install to test it?

thanks!

jakaskerl

Hi ClintCrowe
Run the demo script with python3 depthai_demo.py.

ClintCrowe What else can I be running with just the base install to test it?

Usually you would have a respository called depthai-python (you can manually download it). Inside, there are example scripts and some other utilities you can use to start off.

Let me know if you need anything.

Thanks,
Jaka

ClintCrowe

ok, great, found the demo script and was able to run that to get the camera to work.

how to figure out how to actually us the ai part of it… :-)

thanks!

jakaskerl

Hi ClintCrowe
I suggest going through depthai API docs; mainly, the neural network node and it's yolo and mobilenet versions (you can find them here under nodes. Check the input and output and make sure you fully understand what the network expects you to do. I personally wouldn't first start with standard NeuralNetwork node as it can get confusing with FP layers and result parsing.

We have many examples that pertain to the NN (find them under examples). Something like the yolo example would be a good starting ground.

Of course make sure to understand the framework of the model you are using. These are available all over the internet or you can watch some lessons on YouTube as well.

Let me know if you have any more questions.

Thanks,
Jaka

ClintCrowe

Thanks Jaka.

One of the issues that I am trying to get my head around is the value of the oak-d and the use cases for the greenhouse. I am capturing images with it and placing them in a googledrive folder. Then I run tensorflow on another machine to measure the track growth rate. In that case, the oak-d just seems to be a glorified camera. I'm not sure how it can help me with growth rate.

I guess its more designed to track things like pests? Or disease growth? Something where it is always on and recording information that it can process. Would that be a valid assumption?

thanks!

jakaskerl

Hi ClintCrowe
The OAKs are edge devices that thrive in environments where there are time critical operations (robotics, surveillance, self-driving, ...). In your case of analyzing slow moving systems like growing plants while also not using depth imaging or AI vision, the OAK-D really would be just a glorified camera.
The point of OAKs is offloading all processing to the device itself so it can be used in variety of use-cases where fast PCs would be logistically impossible or too power inefficient.

Thanks,
Jaka