How do I get color output from appsink with opencv?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Currently I am using this pipeline to get video frames in appsink:

uridecodebin uri="file:///home/stiv/lpr/data/dual_day.avi" !nvvideoconvert ! appsink

VideoCapture cap(pipe);

    if(!cap.isOpened()){
        cout << "Error opening video stream or file" << endl;
        return -1;
    }
    while(1){
        Mat frame;
        cap >> frame;
        if (frame.empty())
            break;
        cout << frame.channels()<<endl;
        //resize(frame,frame,Size(320,200));
        imshow( "Frame", frame );
        char c=(char)waitKey(25);
        if(c==27)
            break;
    }
    cap.release();
    destroyAllWindows();

As a result I get frame with 1 channel, but it works. How do I get colored image, as it appears in input stream ? (If I put EGL sink in the end, I see color videostream).

2

Hi,
It should work by sending BGR buffers to appsink:

uridecodebin uri="file:///home/stiv/lpr/data/dual_day.avi" ! nvvideoconvert ! video/x-raw,format=BGRx ! videoconvert ! video/x-raw,format=BGR ! appsink

You may try I420 also. Please refer to the sample in
https://devtalk.nvidia.com/default/topic/1042892/jetson-agx-xavier/nvarguscamerasrc-opencv-solved-/post/5290833/#5290833

3

UPDATE: finally it worked:

uridecodebin uri="file:///home/stiv/lpr/data/dual_day.avi" ! nvvideoconvert ! video/x-raw,format=(string)BGRx ! videoconvert ! video/x-raw,format=(string)BGR ! appsink

Thank you so much. It would be cool to understand why I need this line exactly, is it covered anywhere in the documentation?

4

Advice from your second link doesn’t work unfortunately, because it refers to nvvidconv, which is not found. Also tried flip-method=2, and it also doesn’t help.

5

BGR is the main format in OpenCV. Since nvvideoconverter doesn’t support this format, videoconvert is required to convert to BGR.

nvvidconv is on Jetson platforms.

6

hello,when i push a rtsp streaming to gst rtsp server ,can u give me some suggestion about rtsp server launch commnd?

7

Hi,
You can use test-lauch

For using OpenCV, below post may help:
[url]https://devtalk.nvidia.com/default/topic/1064809/jetson-nano/how-to-add-empty-audio-track-to-video-stream/post/5394002/#5394002[/url]

8

thank u

9

But videoconvert is terribly slow (9 fps).
Is there any way to get nvvidconv, or nvvideoconvert to OpenCV appsink?

10

Hi,
For using appsink to get buffer in OpenCV, you have to suffer the extra emmcpy() from NVMM buffer to CPU buffers. A similar discussion is in
https://devtalk.nvidia.com/default/topic/1064944/

You may check if you ca apply your case to deepstream-app to achieve optimal performance.