We can see that there is a YoloDetectionNetwork node to decode the output of yolo.

I have some yolov3/yolov4 models, they are built by different deep learning frameworks(such as pytorch, tensorflow...), and their output is a bit different. I want to use them on Oak, and I want to use the YoloDetectionNetwork node to parse the output.

What I want is what model output the YoloDetectionNetwork node needs, so that I can adjust the existing model to fit the YoloDetectionNetwork node.

  • erik replied to this.

    Hello Mulong ,
    it's the standard YOLO output - so confidence, bounding boxes and class. There are two output layers:

    Layer name: conv2d_17/BiasAdd/Add, Type: DataType.FP16, Dimensions: [1, 255, 13, 13]
Layer name: conv2d_20/BiasAdd/Add, Type: DataType.FP16, Dimensions: [1, 255, 26, 26]

    For decoding, I would look into the official YOLO github repo at examples - how they decode it and display the results.

    Thanks, Erik