twlelev / openvino_cpp_yolox-yolov5-yolov8-yolov9 Goto Github PK

Tried with yolov8n, it's doesn't work the onnx file work with python but after converting it to bin/xml it gives 0 outputs:
I converted it with:
yolo export model=best640.pt opset=12 simplify=True dynamic=False format=onnx imgsz=640
mo --input_model best.onnx -s 255 --reverse_input_channel True --compress_to_fp16=False

I Tried after convert onnx to IR(bin/xml), but i got fasle detections without the real one with yolov4-tiny.
I used yolov8 project for it, it should support it?

it was a great help to me ,i would very much like you to publish content about the bookstore the Openvinov8 sement

terminate called after throwing an instance of 'ov::AssertFailure'
what(): Check 'm_impl->m_outputs.size() == 1' failed at src/core/src/preprocess/pre_post_process.cpp:123:
PrePostProcessor::output() - Model must have exactly one output, got 8

Thanks for your work, I was wondering if the quantized int8 model of nncf could be used

hi, i have bug.

When I use the following code to export the onnx model

model = YOLO("./runs/detect/train3/weights/last.pt")
model.export(format='onnx',dynamic=True)

Code running throws an exception： get_shape was called on a descriptor::Tensor with dynamic shape