How to deploy a .blob file?

AliChoudhry

Hello! New to the world of computing/programming so please excuse me if any information is missing.

So I started with a set of generated images on Roboflow as a train/val/test dataset or annotated images.
These I trained on Ultralytics and then evaluated.
From here, I was able to output to a variety of file formats including .pt, .onnx, and .xml/.bin

I found the .onnx output was the easiet with this code:

`import blobconverter

blob_path = blobconverter.from_onnx(
model="/path/to/model.onnx",
data_type="FP16",
shaves=5,
)`

The model I'm trying to make is a simple object detection model but I'm a little stuck now. What do I do with the .blob file to run it on the Luxonis Oak-1?

Any help would be appreciated.

jakaskerl

Hi AliChoudhry
That depends on the model framework you are using.
We have custom node for both yolo and mobilenetSSD models (built on top of NeuralNetworkNode). If you use a random model, you will need to create a base NN node (https://docs.luxonis.com/projects/api/en/latest/components/nodes/neural_network/#neuralnetwork).
In any case, you will need to specify nn.setBlobPath(BlobPath) so the node knows which blob you want to use.

Examples are available here:
yolo- https://docs.luxonis.com/projects/api/en/latest/samples/Yolo/tiny_yolo/
ssd- https://docs.luxonis.com/projects/api/en/latest/samples/MobileNet/rgb_mobilenet/
random- https://docs.luxonis.com/projects/api/en/latest/samples/NeuralNetwork/concat_multi_input/

Hope this helps,
Jaka