Csicam with nvosd

[ClementDps]

Hello,
I modified the deepstream_test1 in the deepstream 5.0 samples to have as an input the csi camera. It works but the video is very very slow (about 1 FPS). It looks like frames are being processed normally but only the nvosd is being slow.
Here is my pipeline:
nvarguscamera → capsfilter ->nvstreammux → nvinfer ->nvvidconv ->nvosd → nvegltransform → nveglglessink

Do you know why it is being so slow?
Thanks,
Clément

[ClementDps]

I am still having this problem. Is there a sample maybe I could take a look at? Or is there an explanation for this slow on screen display?

[Amycao]

Can you add queues between every two elements after nvstreammux to see if improvements?

[ClementDps]

Solution in here:

[Nano] Deepstream-test3-app not working as expected for multiple video source

Hi, Please try nvoverlaysink: #ifdef PLATFORM_TEGRA - transform = gst_element_factory_make ("nvegltransform", "nvegl-transform"); + transform = gst_element_factory_make ("queue", "queue"); #endif - sink = gst_element_factory_make ("nveglglessink", "nvvideo-renderer"); + sink = gst_element_factory_make ("nvoverlaysink", "nvvideo-renderer"); And set ‘sync=false’ to nvoverlaysink: g_object_set (G_OBJECT(sink), "sync", FALSE, NULL);