Can I access to objects bounding box coordination which tensorrt engine gives to nvinfer. I think only getting bb coordination is possible and setting them manually (or changing some of objects location which trt gives) is impossible. sometimes I need to ignore some objects location tensorrt gives their location to nvinfer. How can I do that?
in our application inference is done for detection no classifiaction.
Hi,
Please check the source below:
You can filter or modify the bounding box before output.
Thanks.
Hi AastaLLL
I don’t find any function to process objects in nvinfer for primary(detection) mode. there is a function with name gst_nvinfer_process_objects which is called only in secondary mode.
Can you show me which function should I change?
Hi AastaLLL
I don’t find any function to process objects in nvinfer for primary(detection) mode. there is a function with name gst_nvinfer_process_objects which is called only in secondary mode.
Can you show me which function should I change?
I found it . It is attach_metadata_detector function.
Hi,
Is your issue fixed after updating the function?
Thanks.
yes.It fixed.
Ty