Hardware Platform: dGPU • TensorRT Version: 8.6.1 • NVIDIA GPU Driver Version: 550 • Issue Type: question
I have been using nvds_obj_enc_process to get encoded object. Here is overview of my pipeline:
detect -> tracker -> recognition(nvds_obj_enc_process here if object is valid and is the first object) -> nvdemux -> nvconvert -> nvosd -> fakesink
I got mismatch frame_meta.frame_num very randomly when i get NVDS_CROP_IMAGE_META on downstream element. For example:
Run enc object process on obj_meta at frame_num=21 at upstream element.
Got encoded object image at obj_meta at frame_num=22 at downstream element.
What can be the cause of this missmatch. As i said it is very random and i dont get this every run.
Can you reproduce this issue by /opt/nvidia/deepstream/deepstream/sources/apps/sample_apps/deepstream-image-meta-test/deepstream_image_meta_test.c ?
Is nvds_obj_enc_process called for every frame? Is encoding triggered by some condition?
No i only called if it is the first object provided by the tracker or when new object has higher confidence. And it seems this problem only occur on first object.
Can you share sample code and a stream that reproduces the issue?
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
system
August 14, 2025, 3:38am
7
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.