In the deepstream-test-5 example, I want to receive RTSP streams using nvvideoconvert and crop them with croplay. I understand that I need to use the ‘src-crop’ property, but I’m not sure how to use it. Should I configure this in the config file? Can you help me with this?
PS: this is my first topic in here, if I wrong something please teach me :)
This may need to modify our source code. Could you attach your pipeline graph by referring to our FAQ first? Then we can find the position in the pipeline where you want to crop. Finally, we can set the “src-crop” parameters through the nvvideoconvert plugin founded.
Actually, I don’t have much knowledge about pipelines yet; I’m just trying to learn while making progress by compiling on test-5. What I want to do is to receive the RTSP stream as a single segment. After receiving it as a single segment, I want to crop it into two or three areas and send it to inference just before that step. After inference, I want to combine it back into a single stream and use smart record.
The reason I want to do this is that I know my model performs much better with square inputs. By sending the incoming stream to inference as square segments, I aim to use my model with maximum performance.
We do not currently support this operation. If you just need to improve the performace of your model, you don’t need to crop it. We will scale the image according to the input of your model.
There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks
I changed my development approach, built a new pipeline from scratch, and achieved my desired result by adding an nvvideoconvert exactly where and how I wanted. Thank you for your help.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.