Interesting project but how's the latency you tested ?

[Stephenfang51]

Thanks for the cpp work for YOLOP
Just wonder it needs low-latency, high fps for the task such as autonomous driving.
Could an onnx-model loaded through OpenCV use CUDA kernel?

Have you test the latency for the inference time?
Thanks !