Depstream decoded data

For help, how can I get every frame of data decoded by deepstream?

Hi,
Please leverage dsexample plugin:

deepstream_sdk_v4.0_jetson\sources\gst-plugins\gst-dsexample\README

https://docs.nvidia.com/metropolis/deepstream/4.0/dev-guide/index.html#page/DeepStream_Development_Guide%2Fdeepstream_custom_plugin.html

Hi, DaneLLL
When I compiled, the following error occurred:
Makefile:25: *** “CUDA_VER is not set”. Stop.
How can I solve it?

Hi,
You can set it along with ‘make’

$ CUDA_VER=10.0 make

Hi,DaneLLL
Whether the data decoded by deepstream is NV12 or not, can I convert it to BGR format through the function get_converted_mat in gst_dsexample?

Hi,
Please check this sample to use cv APIs.

cv::cvtColor (in_mat, out_mat, CV_RGBA2BGR);
static gint dump = 0;
if (dump < 150) {
    char filename[64];
    snprintf(filename, 64, "/home/nvidia/image%03d.jpg", dump);
    cv::imwrite(filename, out_mat);
    dump++;
}

Hi,DaneLLL
But I just want to get the original data decoded by deepstream. I don’t want multi-channel mixing and object detection. What should I do?How can I change deepstream-app?

Hi,
If you don’t need to run deeplearning models, you may refer to gstreamer guidance for pure video decoding:
https://developer.nvidia.com/embedded/dlc/l4t-accelerated-gstreamer-guide-32-2

Hi,DaneLLL
I see that the deepstream-app instance will detect people or cars. If I want to get pictures of people or cars being tested, what should I do?

Hi,
If you don’t need object-detection, you may not use Deepstream SDK and better to check accelerated gstreamer user guide.

A question about setting CUDA_VER to compile gst-dsexample.
When I set CUDA_VER=10.0 via export CUDA_VER=10.0 then sudo make install fails saying that CUDA_VER is not set.
However if I run sudo CUDA_VER=10.0 make install it works just fine??

Hi,
You should enter super user and then do ‘export CUDA_VER=10.0’

nvidia@nvidia-desktop:~/deepstream_sdk_v4.0.1_jetson/sources/gst-plugins/gst-dsexample$ sudo su
root@nvidia-desktop:/home/nvidia/deepstream_sdk_v4.0.1_jetson/sources/gst-plugins/gst-dsexample# export CUDA_VER=10.0
root@nvidia-desktop:/home/nvidia/deepstream_sdk_v4.0.1_jetson/sources/gst-plugins/gst-dsexample# make install
cp -rv libnvdsgst_dsexample.so /opt/nvidia/deepstream/deepstream-4.0/lib/gst-plugins/
'libnvdsgst_dsexample.so' -> '/opt/nvidia/deepstream/deepstream-4.0/lib/gst-plugins/libnvdsgst_dsexample.so'
root@nvidia-desktop:/home/nvidia/deepstream_sdk_v4.0.1_jetson/sources/gst-plugins/gst-dsexample#