Hello Experts,
I have an IP camera streaming data in RTSP with H.265 encoding. Is it efficient (does it use GPU) to run opencv python lib with default CNN model on face recognition in received video.
Where can I find the source for opencv and CUDA core interaction library source
DaneLLL
October 5, 2020, 12:39am
#3
Hi,
Since OpenCV is CPU-based software frameworks, we would suggest not to use it. Please try DeepStream SDK.
Key feature highlights
• New inference capability with Triton Inference Server enables developers to deploy models natively in their training framework.
• Choice of developing DeepStream pipeline from either C/C++ or Python with comparable performance
• Build and deploy DeepStream apps natively in Red Hat Enterprise Linux (RHEL)
• Enhanced IoT capabilities, including edge/cloud bi-direciotnal messaging and OTA updates
Read more about the announcement and full list of latest features in our …
After the installation through SDKManager, you will see the samples in
/opt/nvidia/deepstream/deepstream-5.0/
Please take a look and give it a try.