OpenCV DNN not making use of Jetson Nano GPU

I have installed Opencv4.0.0 from the script and I want to run yolov3 darknet on GPU of Jetson nano using opencv DNN. Right now its running on CPU and making use of all 4 cores and time taken is 5 sec per frame. If its not possible to run to GPU whats the other alternative?

Hi, it would appear that OpenCV DNN module doesn’t support CUDA or cuDNN.

To get the best performance, we recommend running YOLO with TensorRT. See the instructions here to run it and the code:

[url]https://devtalk.nvidia.com/default/topic/1050377/jetson-nano/deep-learning-inference-benchmarking-instructions/[/url]

Recently, dnn module supports CUDA backend.
https://github.com/opencv/opencv/pull/14827