sei23
January 18, 2023, 6:44am
1
• Hardware Platform (Jetson / GPU) Jetson
• DeepStream Version 6.1
• JetPack Version (valid for Jetson only) Unknown
• TensorRT Version 8.4
• Issue Type( questions, new requirements, bugs) questions
Hello.
I’m using PeopleSegNet for inference and I want to retrieve NVDSINFER_SEGMENTATION_META in my own gstreamer plugin.
I refer to dsexample source code and my current implementation is
frame_meta = (NvDsFrameMeta *)(l_frame->data);
for (l_user = frame_meta->frame_user_meta_list; l_user != NULL;
l_user = l_user->next)
{
user_meta = (NvDsUserMeta *)(l_user->data);
g_print("************ meta_type is %d\n", user_meta->base_meta.meta_type);
if (user_meta->base_meta.meta_type == NVDSINFER_SEGMENTATION_META)
{
g_print("NVDSINFER_SEGMENTATION_META\n");
}
}
This code works without any errors but it doesn’t print any message I set above.
I assume I’m doing wrong.
Could you give me any suggestions?
fanzh
January 18, 2023, 9:01am
3
please refer to PeopleSegNet deepstream demo GitHub - NVIDIA-AI-IOT/deepstream_tao_apps: Sample apps to demonstrate how to deploy models trained with TAO on DeepStream , from the cfg pgie_peopleSegNet_tao_config.txt, you need to set output-instance-mask=1, then Gst-nvinfer attaches instance mask output in object metadata, please refer to attach_metadata_detector of gst-nvinfer\gstnvinfer_meta_utils.cpp. so mask data is in object 's obj_meta->mask_params.data.
sei23
January 19, 2023, 1:22am
4
I’ve already enabled output-instance-mask flag and I was able to retrieve mask from mask_params.data.
But my question is
How can i retrieve data from frame_user_meta_list?
I think nvinfer docs say it attaches mask data to frame_user_meta_list.
fanzh
January 19, 2023, 1:28am
5
the nvinfer doc said “Gst-nvinfer attaches instance mask output in object metadata.”, it is in object metadata, frame_user_meta_list is used to save user meta, object meta and user meta are different, please refer to MetaData in the DeepStream SDK — DeepStream 6.1.1 Release documentation
system
Closed
February 2, 2023, 1:30am
7
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.