VideoProcessingFramework: Both nvDec.DecodeSingleSurface() and nvDec.DecodeSingleFrame() hang on RTSP from specific camera just after opening

Hi!

Awesome piece of software, thanks for your effort!

Here is my problem: Trying to read from an rtsp source corresponding to an Axis model camera (URL: rtsp://[user:passwd@IP]/axis-media/media.amp) with a code directly based on the examples (see below)…

    [...]
    nvDec = nvc.PyNvDecoder(encFilePath, gpuID, {'rtsp_transport': 'tcp', 'max_delay': '5000000', 'bufsize': '300000k'})
    width, height = nvDec.Width(), nvDec.Height()
    print(width,height)
    while True:
        try:
            print("YES")
            rawSurface = nvDec.DecodeSingleSurface()
            print("NO")
         [...]

… the method nvDec.DecodeSingleSurface() never returns. The same occurs when trying nvDec.DecodeSingleFrame().

Otherwise, trying the exact same piece of code with other two different cameras (Hikvision and Reolink models; URLs rtsp://[user:passwd@IP]/h264Preview_01_main and rtsp://[user:passwd@IP]/h265/ch1/main/av_stream), the program works seamlessly, correctly returning the read frames in both cameras, both in the reading to RAM and VRAM cases. Note also that the first URL uses h264, while the second uses h265, again both working OK.

FFmpeg (called as usual from the command line) reads also frames correctly in all cases (Axis camera included).

Any ideas on what could be the reason? The Axis camera never gave us a problem, it has been tested with success even using deepstream (based on gstream) with CUDA accelerated decoding. Maybe some rtsp dict parameter that could help? Any clue would be appreciated!

• OS: Linux (Ubuntu 18.04)
• Nvidia driver version: 440.33.01
• CUDA Version: 10.1
• Video Codec SDK Version: 9.1.23
• Python Version: 3.7
• FFMPEG version: 4.3, compiled from source just as explained in the wiki

Thanks in advance, best regards, and I wait for your reply, Pedro E.