How to use deepstream_imagedata-multistream in custom app

Hi all,
I want to use deepstream_imagedata-multistream for decoding of multi-stream, and then use the decoded frames in my custom python app, I want to integrate decoder(deepstream_imagedata-multistream) with my processing part and both work next to each other .

Do you mean to process the decoded video data(in YUV or RGB format)? The sample app deepstream_imagedata-multistream is not just a decoder. There is no interface for app to access the data buffer directly now.
In c level, you can develop your own gstreamer plugin to process frame data.

Yes I want to process the decoded streams in python.
I cant instread of write to disk the decoded frames, push them into queue?

Python interface is application level interface, it can not access the frame data directly.

So there is no way to use decoded frames in custom python app?

I used gstreamer + opencv in python, but this use both GPU buufer + CPU buffer,

        gstream_elemets = (
            'rtspsrc location=rtsp latency=300 !'
            'queue !'
            'rtph264depay ! h264parse !'
            'omxh264dec  disable-dvfs=1 enable-low-outbuffer=1 !'
            'queue !'
            'video/x-raw(memory:NVMM), framerate=25/1, format=(string)NV12 !'
            'queue !'
            'nvvidconv ! video/x-raw, width={}, height={}, framerate=25/1, format=(string)BGRx !'
            'videoconvert ! video/x-raw, framerate=25/1, format=(string)BGR !'
            'appsink max-buffer=100 sync=0 ')

cv2.VideoCapture(gstream_elemets, cv2.CAP_GSTREAMER)

I want to skip some frames, becuasethe input frame rate = 25 and output frame rate = 5, then I don’t want to decode all evey 25 input frames, I want to decode every 5th frames. This option is exist in nvv4l2decoder , but there is not such a way in omxh264dec to skip some frames. and I tested the nvv4l2decder, but I don’t know why this decoder don’t work in rtsp source but work for file source.
I also used videorate elemnt for framerate=5/1, but this cause cpu usage dramatically increased. I have to decoder every 25 frames then take only 5 frame with counter state in python program. because of I don’t want to decode every 25 frames.

please see this link.

Do you mean you want to use “drop-frame-interval” property of nvv4l2decoder to change output framerate?

I don’t want to use “drop-frame-interval” property of nvv4l2decoder for setting framerate of output, But because the output frame rate is lower than input frame rate, why I do doing decoder for all frames, I want to drops some of frames without decoding.

Dropping frame function needs to consider many conditions in pipeline. It is better to implement in c level than in python.