Recherche avancée

Sur d’autres sites (7230)

• Optimizing conversion of video files to texture-atlas

  7 novembre 2017, par Malu05

  I am working on some code that converts a video file into a 1d texture atlas (a single JPG file containing all of the frames stacked on top of each other).
  I do this by reading in the frames using FFMPEG’s image2pipe, and read the pipe into a 3d numpy array. Then use numpy’s concatenate function to add them on top of each other.
  The code itself works fine, but it is somewhat slow. In this example 100 frames of a video takes roughly 15-20 seconds.

  Reading the frames into the pipe is somewhat fast, but reading from the pipe, converting to array and concatenating takes some time.
  I am not too familiar with ways to optimize numpy array handling, but
  I expect that I might be doing something very inefficient here but i can’t seem to spot it. Is there anything i can do to make this process faster ?

  import subprocess as sp
  
  import numpy
  
  import datetime
  
  import matplotlib.pyplot as plt #Only for test plotting.
  
  
  
  FFMPEG_BIN = "ffmpeg" #link to the ffmpeg executeable
  
  
  
  
  
  #Number of images in the texture-atlas
  
  images = 100
  
  
  
  for x in range(0,images): #Loop for each frame to read
  
      res = [596,336,3] #The output format
  
  
  
      #Setup the command for reading 1 frame in the video.
  
      command = [ FFMPEG_BIN,
  
                  '-ss', '%s'%(x),
  
                  '-i', '340114063.mp4',
  
                  '-f', 'image2pipe',
  
                  '-vframes', '1',
  
                  '-vf','scale=%s:%s'%(res[0],res[1]),
  
                  '-pix_fmt', 'rgb24',
  
                  '-vcodec', 'rawvideo', '-']
  
      pipe = sp.Popen(command, stdout = sp.PIPE,stderr=sp.PIPE, bufsize=-1)
  
  
  
      #Read out the image data from the pipe
  
      raw_image = pipe.stdout.read(res[0]*res[1]*res[2])
  
      #Convert to numpy array
  
      image =  numpy.fromstring(raw_image, dtype='uint8')
  
      #Clear the pipe
  
      pipe.stdout.flush()
  
      if x>0:
  
          #Add the image below the last image(s) so that they form a 1d texture atlas
  
          image = numpy.concatenate((last_image,image), axis=0)    
  
      last_image = image
  
  
  
  
  
  #not needed, just for plotting example.
  
  image = image.reshape((res[1]*images,res[0],res[2]))
  
  imgplot = plt.imshow(image)
  
  plt.savefig('foo.png')

• avfilter/ddagrab : create secondary xor mouse texture

  8 octobre 2023, par Timo Rothenpieler

  avfilter/ddagrab : create secondary xor mouse texture
  

  • [DH] libavfilter/vsrc_ddagrab.c

• avfilter/ddagrab : only use acquired texture on valid updates

  9 février 2024, par Timo Rothenpieler

  avfilter/ddagrab : only use acquired texture on valid updates
  

  • [DH] libavfilter/vsrc_ddagrab.c