Recherche avancée

Sur d’autres sites (13346)

• RTSP stream to ffmpeg problems

  14 octobre 2022, par maeek

  I'm writing a web application for managing and viewing streams from ONVIF ip-cameras.
  
It's written in nodejs. The idea is to run a child process in node and pipe output to node, then send the buffer to client and render it on canvas. I have a working solution for sending data to client and rendering it on canvas using websockets but it only works on one of my cameras.

  


  I own 2 IP cameras and both of them have rtsp server.
  
One of them(let's name it camX) kind of works with this ffmpeg command (sometimes it just stops, maybe due to packet losses) :

  


  ffmpeg -rtsp_transport tcp -re -i  -f mjpeg pipe:1


  


  But the other one(camY) returns Nonmatching transport in server reply and exits.

  


  I discovered that the camY transport is unicast but ffmpeg doesn't support this particular lower_transport as I read on ffmpeg forum.

  


  So I started looking for a solution. My first idea was to use openRTSP which works fine with both streams.
I looked at the documentation and came up with this command :
  
openRTSP -4 -c  | ffmpeg -re -i pipe:0 -f mjpeg pipe:1
  
-4 parameter returns stream to pipe in mp4 format
  
And here's another problem I ran into, ffmpeg returns :

  


  [mov,mp4,m4a,3gp,3g2,mj2 @ 0x559a4b6ba900] moov atom not found  
pipe:0: Invalid data found when processing input


  


  Is there any way to make this work ?
I tried various solutions I found, but none of them worked.

  


  EDIT

  


  As @Gyan suggested I used -i parameter instead of -4 but it didn't solve my problem.

  


  My command :

  


  openRTSP -V -i -c -K  | ffmpeg -loglevel debug -re -i pipe:0 -f mjpeg pipe:1
  
Created receiver for "video/H264" subsession (client ports 49072-49073)
Setup "video/H264" subsession (client ports 49072-49073)
AVIFileSink::setWord(): SeekFile64 failed (err 29)
AVIFileSink::setWord(): SeekFile64 failed (err 29)
AVIFileSink::setWord(): SeekFile64 failed (err 29)
AVIFileSink::setWord(): SeekFile64 failed (err 29)
AVIFileSink::setWord(): SeekFile64 failed (err 29)
AVIFileSink::setWord(): SeekFile64 failed (err 29)
AVIFileSink::setWord(): SeekFile64 failed (err 29)
AVIFileSink::setWord(): SeekFile64 failed (err 29)
AVIFileSink::setWord(): SeekFile64 failed (err 29)
Outputting to the file: "stdout"
[avi @ 0x5612944268c0] Format avi probed with size=2048 and score=100
[avi @ 0x56129442f7a0] use odml:1
Started playing session
Receiving streamed data (signal with "kill -HUP 15028" or "kill -USR1 15028" to terminate)...
^C
[AVIOContext @ 0x56129442f640] Statistics: 16904 bytes read, 0 seeks
pipe:0: Invalid data found when processing input


  


  As you can see openRTSP command return err 29 but in meantime it outputs some data to pipe.
  
When I terminate the command ffmpeg shows that it read some data but couldn't process it.

  


  Here's the function that produces that error :

  


  void AVIFileSink::setWord(unsigned filePosn, unsigned size) {
  do {
    if (SeekFile64(fOutFid, filePosn, SEEK_SET) < 0) break;
    addWord(size);
    if (SeekFile64(fOutFid, 0, SEEK_END) < 0) break; // go back to where we were

    return;
  } while (0);

  // One of the SeekFile64()s failed, probable because we're not a seekable file
  envir() << "AVIFileSink::setWord(): SeekFile64 failed (err "
          << envir().getErrno() << ")\n";
}


  


  In my opinion it looks like it won't be able to seek file because it's a stream not a static file.
  
Any suggestion for a workaround ?

  


• Record video with Xvfb + FFmpeg using Selenium in headless mode

  12 mars 2024, par ifdef14

  I am trying to record video using Selenium in headless mode. I am using Xvfb and FFmpeg bindings for Python. I've already tried :

  


  import subprocess
import threading
import time

from chromedriver_py import binary_path
from selenium import webdriver
from selenium.webdriver.chrome.service import Service
from xvfbwrapper import Xvfb


def record_video(xvfb_width, xvfb_height, xvfb_screen_num):
    subprocess.call(
        [
            'ffmpeg',
            '-f',
            'x11grab',
            '-video_size',
            f'{xvfb_width}x{xvfb_height}',
            '-i',
            xvfb_screen_num,
            '-codec:v',
            'libx264',
            '-r',
            '12',
            'videos/video.mp4',
        ]
    )


with Xvfb() as xvfb:
    '''
    xvfb.xvfb_cmd[1]) returns scren num
    :217295622
    :319294854
    :
    '''
    xvfb_width, xvfb_height, xvfb_screen_num = xvfb.width, xvfb.height, xvfb.xvfb_cmd[1]
    thread = threading.Thread(target=record_video, args=(xvfb_width, xvfb_height, xvfb_screen_num))
    thread.start()
    opts = webdriver.ChromeOptions()
    opts.add_argument('--headless')
    try:
        driver = webdriver.Chrome(service=Service(executable_path=binary_path), options=opts)
    finally:
        driver.close()
        driver.quit()



  


  As much as I understand xvfb.xvfb_cmd[1] returns an information about virtual display isn't it ? When I executed this script, I got the error message :

  


  [x11grab @ 0x5e039cfe2280] Failed to query xcb pointer0.00 bitrate=N/A speed=N/A    
:1379911620: Generic error in an external library


  


  I also tried to use the following commands :

  


  xvfb-run --listen-tcp --server-num 1 --auth-file /tmp/xvfb.auth -s "-ac -screen 0 1920x1080x24" python main.py &

  


  ffmpeg -f x11grab -video_size 1920x1080 -i :1 -codec:v libx264 -r 12 videos/video.mp4

  


  In the commands above, there are used xvfb-run --server-num 1 and ffmpeg -i :1, why ?

  


  Overall, when Selenium is running in the headless mode what's going on behind the scenes ? Is it using virtual display ? If yes, how can I detect display id of this, etc. Am I on the right path ?

  


  I am not using Docker or any kind of virtualization. All kind of tests are running on my local Ubuntu machine.

  


• avcodec/vp9 : fix leaked cbs fragment AVBufferRef

  10 août, par Leo Izen

  avcodec/vp9 : fix leaked cbs fragment AVBufferRef
  When this function returns, the fragment is never reset, so
  
  current_frag->data_ref is never unref-ed, which ends up leaking it.
  
  We call ff_cbs_fragment_reset to release the reference on its buffer.
  Signed-off-by : Leo Izen <leo.izen@gmail.com>
  

  • [DH] libavcodec/vp9.c