Video streaming issue due to not enough CPU power

i am using jetson nano to do realtime object detection and now facing some issues in video streaming from IP camera.

I need to save videos during the realtime detection therefore I need to run both video streaming, video writing and AI detection at the same time.

Currently, video streaming and video writing are done by OpenCV, however, due to limited CPU power(all 4 cores are 100% when doing both tasks together) on jetson nano, doing both together cause latency in video streaming.

how could I shift the load from CPU to GPU? is deepstream tackling this issues?

*I am using python.


Please install DeepStream SDK and try deepstream-app. We have default config file in


You can configure source to RTSP source:
And sink to encode + filesave
deepstream-app is in C code. For python code, please refer to