Recherche avancée

Sur d’autres sites (14186)

• Revision 33afddadb9 : Merge "Add variance based mode/skipping"

  6 août 2013, par Deb Mukherjee

  Changed Paths :
   Modify /vp9/encoder/vp9_block.h

  
   Modify /vp9/encoder/vp9_encodeframe.c

  
   Modify /vp9/encoder/vp9_onyx_if.c

  
   Modify /vp9/encoder/vp9_onyx_int.h

  
   Modify /vp9/encoder/vp9_rdopt.c

  
  
  Merge "Add variance based mode/skipping"

• Revision 0daadeb60c : Enable motion field based mode seach skip This commit allows the encoder to che

  12 août 2014, par Jingning Han

  Changed Paths :
   Modify /vp9/encoder/vp9_rdopt.c

  
   Modify /vp9/encoder/vp9_speed_features.c

  
   Modify /vp9/encoder/vp9_speed_features.h

  
  
  Enable motion field based mode seach skip

  This commit allows the encoder to check the above and left neighbor
  blocks’ reference frames and motion vectors. If they are all
  consistent, skip checking the NEARMV and ZEROMV modes. This is
  enabled in speed 3. The coding performance is improved :

  pedestrian area 1080p at 2000 kbps,
  from 74773 b/f, 41.101 dB, 198064 ms
  to 74795 b/f, 41.099 dB, 193078 ms

  park joy 1080p at 15000 kbps,
  from 290727 b/f, 30.640 dB, 609113 ms
  to 290558 b/f, 30.630 dB, 592815 ms

  Overall compression performance of speed 3 is changed
  derf -0.171%
  stdhd -0.168%

  Change-Id : I8d47dd543a5f90d7a1c583f74035b926b6704b95

• mp4 video written with ffmpeg has different first frame based on total number of frames

  25 janvier 2023, par Nitzan Weissman

  I have a code that writes and reads videos in Python, I'm doing the writing using ffmpeg. I noticed a weird phenomenon, where the first frame changes based on the total number of frames in the video. I find it weird because the first frame is a key frame (I checked it using ffprobe)

  


  The code I'm using :

  


  The writing function

  


  import shutil
import subprocess
def write_h264_video(video_path: str, images: np.ndarray, fps: float):
    image_name_template = '00%04d.png'
    tmp_image_dir = (Path(video_path).parent / 'tmp_imgs')
    tmp_image_dir.mkdir()

    for ii, image in enumerate(images):
        cv2.imwrite(f'{str(tmp_image_dir / str(ii).zfill(6))}.png', image[..., ::-1])    
    command = f'ffmpeg -hide_banner -loglevel error -framerate {fps} -y -i {tmp_image_dir}/{image_name_template} ' \
            f' -c:v libx264 -pix_fmt yuv420p {video_path} -y'
    subprocess.call(command, shell=True)
    shutil.rmtree(tmp_image_dir)


  


  The main code

  


  import numpy as np
from pathlib import Path
import decord
decord.bridge.set_bridge('torch')

input_video_path = Path('')
tmp_path = Path('')
input_video = decord.VideoReader(str(input_video_path))

frames = np.zeros((20, 512, 512, 3), dtype = np.uint8)
for i in range(20):
    frames[i] = input_video[i].detach().cpu().numpy()
    
long_video_path = tmp_path / 'long_video.mp4'
short_video_path = tmp_path / 'short_video.mp4'

fps = 30
write_h264_video(str(long_video_path), frames, fps=fps)
write_h264_video(str(short_video_path), frames[:10], fps=fps)


long_video = decord.VideoReader(str(long_video_path))
short_video = decord.VideoReader(str(short_video_path))

long_frame_0 = long_video[0]
short_frame_0 = short_video[0]
(long_frame_0 == short_frame_0).all()


  


  I am supposed to get True, but I'm getting False.
I tried reading using other methods, using opencv for example, and the issue persists.
There are other ways to write videos of course, but I'm interested in ffmpeg-based ways.

  


  Thanks !