Getting a gst-pipeline into detectnet in python


after managing the detection I wanted on the commandline with gst-launch + detectnet I am now trying to get this done in Python.
using a cv2.VideoCapture with the appropriate queue I get the pictures from my IP-camera. imwrite puts a proper jpeg on Disk. However I want to put the frames into the Detect method of jetson.inference.Detectnet, so I am guessing I either need to convert the frame I get from or I shall use a different tooling instead of a cv2.VideoCapture to place the gst-queue in. What is the proper way?



Hi @konstantin3, please see this example for converting OpenCV image to CUDA image using jetson.utils.cudaFromNumpy():