yolov8 vs yolov5 work on barracuda

Unity Engine
, , , ,
1

I discovered that YOLOv8 and YOLOv5 perform instance segmentation quite effectively in photos. However, I couldn’t find any resources explaining how to implement the YOLOv8 model specifically in Unity using Barracuda. To make the best use of the model in Unity, i am wondering whether should i choose YOLOv8 or YOLOv5. And any implementation tutorial.

Here’s my workflow:

  • find model in the roboflow
  • train at colab
  • export to ONNX format
  • adapt it into unity barracuda
    the hardest point is the Barracuda keep throw me errors about the yolov8 model i converted. And here’s my script for convertion:
    /////////////////////////////
    from ultralytics import YOLO

Load a model

model = YOLO(“best.pt”) # load a pretrained model (recommended for training)

Use the model

path = model.export(format=“onnx”) # export the model to ONNX format
//////////////////////////////

2

Setting parameter opset to 10 , can fix the issue of load.
opset versions link Compatibility | onnxruntime

3

Hi… setting the opset to 10 helped in loading the model. however, when running a model that is converted from Yolov8 it gives a Rechape error. but when running a model that is converted from Yolox it works perfectly fine on the same project. So, what could be the issue?

4

Hi,
I’ve converted the YOLOv8n model to ONNX by using
“yolov8n.pt” instead of “best.pt” as I was getting an error when converting “best.pt”. I dragged and dropped this model to the Editor. It did not generate any errors. Could you upload your converted ONNX model together with the full error message you get and the exact steps it took to reproduce the error (with your uploaded model)?

You can use SendSafely to upload your model: Unity

Thanks.