Changing the bounding boxes before secondary classifier

Hello, I’m working on a deepstream app and this is the basic structure:

source → primary-object-detecor-> tracker → secondary-classifier → sink

Is it possible to change the bounding box size of an object (reducing the width by 50px for example) before passing it to the secondary classifier?

Yes, I think you can do it via installing a probe on the sink pad of sgie, you can refer deepstream_tlt_apps/apps/tlt_others/deepstream-faciallandmark-app at release/tlt3.0 · NVIDIA-AI-IOT/deeps