DetectNet: Crop image with detected object(bbox) and run a second detection on it

I am using Jetson-Inference and implemented two detection networks in c++. The first one detects a person and the second one a hand pose. It works great but I only wanna detect a hand pose of the detected person. The idea is that I crop the image to the detected bbox and create a new image in memory which I pass to the hand pose detector. However, I am struggling with the implementation. Basically, I have this uchar3* image pointer to the image. But how can I crop it? Are the any handy helpers available?

Hi @sbrockmann, see examples of using the cudaCrop() function here:

1 Like

Amazing!!! That’s what I was looking for. Thanks a lot :-)

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.