Recherche avancée

Sur d’autres sites (14849)

• How to Capture/Convert video using VLC or FFmpeg starting at the next encountered I-Frame

  28 juillet 2014, par DavidG

  I am capturing multicast video using VLC.exe and performing some processing/converting using FFmpeg.exe. The resulting video starts with a gray background with some of the moving objects moving through the gray background. This seems to me that the video is not starting on an I-Frame, but on a B-Frame or a P-Frame. However, when I run ffprobe, it states the first frame is a key frame. After about 1/2 second, the video plays fine.

  Here is the command I’m using to capture the video :

  vlc.exe -I curses --run-time=10 "my.sdp" --sout=video.mp4

  I’m not using anything special when I convert to an mp4 file using FFmpeg :

  ffmpeg.exe -i video.mp4 -s 400x300 final.mp4

  I tried to post an image of the first frame or two, but Stackoverflow gave me an error stating I need at least 10 reputation to post images. Sorry about that, an image would have helped.

  I ran the command "ffprobe show_frames " and it looks like the first frame is a key frame. Of course I’m new to all this and I don’t know what I’m doing or seeing.

  Here is beginning of the output to the "ffprobe show_frames " command :

  [FRAME]
  
  media_type=video
  
  key_frame=1
  
  pkt_pts=0
  
  pkt_pts_time=0.000000
  
  pkt_dts=0
  
  pkt_dts_time=0.000000
  
  pkt_duration=1001
  
  pkt_duration_time=0.033367
  
  pkt_pos=48
  
  pkt_size=756
  
  width=400
  
  height=300
  
  pix_fmt=yuv420p
  
  sample_aspect_ratio=1:1
  
  pict_type=I
  
  coded_picture_number=0
  
  display_picture_number=0
  
  interlaced_frame=0
  
  top_field_first=0
  
  repeat_pict=0
  
  reference=3
  
  [/FRAME]

  Is there any way to either tell VLC.exe to start capturing on an I-Frame (This would be the best solution as I specify a duration to capture the video) or is there a way to tell FFmpeg.exe to start converting the input video starting at the first I-Frame encountered ?

  Thanks,
   David

• FFMPEG-WEBM : Issue with premultiplied alphas WebM

  17 septembre 2017, par Blake Buell

  I rendered a transparent Mov with a semi-transparent logo on it in Vegas Pro 15 and I converted that to a WebM using FFmpeg so I can put it on my website. When I looked at the logo with the WebM in chrome it had this gray tint compared to the original that was a lot brighter. I’m having a feeling this has something to do with premultiplied alphas but I don’t know how to deal with this with a WebM on chrome.
  I think I would need to convert the premultiplied alpha to a straight alpha and if so how would I do this with FFmpeg ?

  Here is the command I used to convert to WebM if this is useful.

  ffmpeg -i temp/Digi-Logo.mov -c:v libvpx -b:v 2000k -cpu-used 2 -pix_fmt yuva420p -auto-alt-ref 0 -deadline 600 logo/1.webm

  Original :
  

  Converted :
  

  EDIT : My guess is that FFmpeg when converting the MOV to WebM it removes the premultiplied alpha channel data and makes it that gray color. How would I make it so FFmpeg converts the transparent MOV to WebM and keeps the premultiplied alpha channel.

• How to extract frames from a mpeg-2 file in Python ?

  14 avril 2016, par Agnirudra

  I am trying to read and extract frames (faces detection to be precise) from a video file using python. My code is working fine with extracting frames and detecting the faces from a mp4 video file, but it can’t read a mpeg2 video file. What should be the proper approach to fix this problem ?

  import pylab
  
  import imageio
  
  import cv2
  
  import sys
  
  filename = 'c:\file.mp4'
  
  vid = imageio.get_reader("file.mp4",  'ffmpeg')
  
  nums = [10,100,190,250, 150]
  
  for num in nums:
  
      image = vid.get_data(num)
  
      fig = pylab.figure()
  
      fig.suptitle('image #{}'.format(num), fontsize=20)
  
      pylab.imshow(image)
  
  pylab.show()
  
  pic=cv2.imwrite('messigray.png',image)
  
  
  
  
  
  
  
  # Create the haar cascade
  
  faceCascade = cv2.CascadeClassifier('C:\opencv\sources\data\haarcascades\haarcascade_frontalface_default.xml')
  
  # Read the image
  
  image = cv2.imread('C:/Users/agni/Documents/Python_Devs/messigray.png')
  
  gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
  
  # Detect faces in the image
  
  faces = faceCascade.detectMultiScale(
  
      gray,
  
      scaleFactor=1.1,
  
      minNeighbors=5,
  
      minSize=(30, 30),
  
      flags = cv2.cv.CV_HAAR_SCALE_IMAGE
  
                                      )
  
  print "Found {0} faces!".format(len(faces))
  
  
  
  # Draw a rectangle around the faces
  
  for (x, y, w, h) in faces:
  
      cv2.rectangle(image, (x, y), (x+w, y+h), (0, 255, 0), 2)
  
      cv2.imshow("Faces found" ,image)
  
      cv2.waitKey(0)