I am getting this error message while converting my custom yolov3 .xml and .bin file to Blob, so that i can run on OAK-d kit. kindly help
code:
import blobconverter
blob_path = blobconverter.from_openvino(
xml=xmlfile,
bin=binfile,
data_type="FP16",
shaves = 5)
error message attached below.

  • erik replied to this.

    Hello HaroonAsad , we apologize for the inconvenience. I believe you have the old version of blobconverter that is deprecated. Could you try checking your current version of the blobconverter (pip show blobconverter)?
    Thanks, Erik

      yes i have checked the version its given below in snapshot. if its the version which is no longer available
      then guide me to correct it. Thanks

        HaroonAsad Could you check again if your conversion works? I made some changes to blobconverter cloud configuration and managed to successfully convert the NN created in this notebook

          thank you so much, i have got the blob file for my custom yolo object detector but when i run this blob file it do not show any detection and also show frame rate 1 FPS. can you guide me with this error too? Thankyou

          that's the error am getting also its not detecting my custom object. kindly guide

          • erik replied to this.

            Hello HaroonAsad, as per error message, you are using NN model that accepts 416x416 frames, but you are feeding it 300x300 frames. If you linked camRgb.preview to the input of NN node, you can change preview frame size with camRgb.setPreviewSize(416, 416). Example code here.
            Thanks, Erik

              yes i have tried using 416x416 frames but still it did not detected anything. i wanna ask that is OAK-D kit supports only tiny versions of YOLO or main versions of YOLO too?

              • erik replied to this.

                HaroonAsad accuracy of tiny yolo isn't as good, but should detect people without a problem (since it was trained on many images of people). You can use full sized yolo as well, but it's a huge network that will run at 2FPS on OAK. It supported out-of-the-box with our depthai_demo using python3 depthai_demo.py -cnn yolo-v3 or python3 depthai_demo.py -cnn tiny-yolo-v3.
                Thanks, Erik