Recherche avancée

Sur d’autres sites (8707)

• What determine number of cpu core usage of filter select=gt(scene,0.1) ?

  15 décembre 2022, par kocoten1992

  I've notice that when using filter gt(scene,0.1), for example :

  


  ffmpeg -i big_buck_bunny.mkv -filter:v "select='gt(scene,0.1)',showinfo" -f null -

  


  Depends on the video, number of cpu cores usage varies extremely (sometimes it 3 cores usage - other time 12 cores usage in different video).

  


  Would like to ask what determine that logic ?

  


  I try to read ffmpeg source code but not familiar with it, a general explanation would be enough, but much appreciate if you point out the line/directory determine that logic in https://github.com/FFmpeg/FFmpeg.

  


  (Also not asking how to reduce cpu usage, interested in the logic determine that).

  


• OpenCV reading from live camera creates a short video that moves quickly

  17 novembre 2022, par user19019404

  I am reading in a live vide stream from a CCTV camera. The camera is set to 5 fps, another is set to 25fps and another to 30fps. Irrespective of the FPS that the camera is set, I can record 5 minutes but end up with a 30 second recorded clip where everyone is running around the scene.

  


  My code is the 'typical' read in video and write video code that you would find online such as (code below simplified for readability) :

  


  import cv2

video = cv2.VideoCapture(live RTSP address of camera)

if (video.isOpened() == False):
    print("Error reading video file")
else:
    frame_width = video.get(cv2.CAP_PROP_FRAME_WIDTH)
    frame_height = video.get(cv2.CAP_PROP_FRAME_HEIGHT)
    frame_fps = video.get(cv2.CAP_PROP_FPS)
    size = (frame_width, frame_height)
    result = cv2.VideoWriter('filename.avi',cv2.VideoWriter_fourcc(*'MJPG'),frame_fps , size)

    while(True):
        ret, frame = video.read()
        if ret == True:
            result.write(frame)
            cv2.imshow('Frame', frame)
        if cv2.waitKey(1) & 0xFF == ord('s'):
            break
        else:
            break
    video.release()
    result.release()
    cv2.destroyAllWindows()
print("The video was successfully saved with new fps")


  


  I have tried playing with the FPS by reading in the FPS from the live camera and using the same FPS in the video write, but all that results is a video that is a fraction of the real time and with people zooming around the scene. So watching a 5 minute smooth video results in a 20 second recorded video with everyone zooming around.

  


  Is this something that I need to fix in the writing of the video or do I need a second pass with ffmpeg to readjust the video ?

  


  Much appreciated

  


  Update, corrected the code above and :
When printing the frames read and the frame written the numbers are the same, showing that each frame that is read is being written (so I am not losing frames along the way thereby writing half the amount of frames).

  


• How to add two video filters in one command line which has transcoding done by two GPUs ?

  3 août 2022, par Max Dax

  I have two video filters :

  


  


  -vf "hwdownload, drawtext=fontfile=/usr/share/fonts/TTF/Verdana.ttf:text='Hello World',
format=nv12, hwupload"

  


  


  and

  


  


  -vf "select='gt(scene,0.04)'" -vsync 0 'snapshot.png'

  


  


  How can I add these two video filters to the below command line which does transcoding from a webcam livestream of MJPG to HEVC using two GPUs ? :

  


  


  ffmpeg
-init_hw_device cuda=decdev :/dev/dri/renderD129
-init_hw_device vaapi=encdev :/dev/dri/renderD128
-hwaccel cuda
-hwaccel_device decdev
-hwaccel_output_format cuda
-c:v mjpeg_cuvid
-f v4l2 -input_format mjpeg -framerate 30 -video_size 1920x1080
-i /dev/video0
-filter_hw_device encdev
-c:v hevc_vaapi webcam.mp4"

  


  


  Basically what I am trying to do is using FFMPEG to transcode and record a live webcam stream and also at the same time take snapshots when motion is detected.

  


  Thanks