for determined part of video


I have a video file, I can run on it to do object classification.

I have radar datas related to video, so I know where is the object what I want to identify by My target is to do the identification more effective way by examine just a part of every video frame.

Question 1: Is there any way to run just a determined part of the video?

Question 2: Does the whole procedure make sense? Does the procedure more effective if the size of the cropped video frames are 1/5 of the original size?

Thank you

Hi @feketegyerek, you could use the cudaCrop() function to crop the image to your region-of-interest (ROI). For more info, see these links:

Hi, thank you for your response!

Currently, my program crops every images with openCV and I run detectNet on these frames. It works fine, but I would like to skip this step, and tell the detectNet where is the object on the whole picture to examine just that part of the image. I don’t want to crop and save sub-images at all.

Is it possible?

Hi @feketegyerek, detectNet does not do the cropping, so you would need to do that beforehand (either with CUDA or OpenCV).