Recherche avancée

Sur d’autres sites (10432)

• How to avoid color shift using FFmpeg libx264

  7 septembre 2021, par florit

  I’d like to convert some footage to proxy H.264 using ffmpeg. I got a slight colorshift using the following command :

  


  ffmpeg -i /path/to/045_0054.MXF -map 0 -dn -c:v libx264 -preset slow -profile:v main -level:v 4.1 -coder 1 -pix_fmt yuv420p -crf 25 -vf "scale=1280:720:flags=lanczos" -movflags +faststart -colorspace 1 -color_primaries 1 -color_trc 1 -color_range 1 -brand mp42 -refs 1 -x264-params b-pyramid=0 -bf 2 -g 5 -y /path/to/045_0054.mp4


  


  How can I get rid of the slight magenta shift ?

  


  mediainfo prints the following about the source file :

  


  General
Complete name                            : /path/to/045_0054.MXF
Format                                   : MXF
Format version                           : 1.3
Format profile                           : OP-1a
Format settings                          : Closed / Complete
File size                                : 180 MiB
Duration                                 : 14 s 720 ms
Overall bit rate mode                    : Variable
Overall bit rate                         : 102 Mb/s
Encoded date                             : 2021-02-19 13:13:15.000
Writing application                      : Sony Mem 2.00

Video
ID                                       : 2
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L5.1
Format settings                          : CABAC / 2 Ref Frames
Format settings, CABAC                   : Yes
Format settings, Reference frames        : 2 frames
Format settings, GOP                     : M=3, N=6
Format settings, wrapping mode           : Frame
Codec ID                                 : 0D01030102106001-0401020201314001
Duration                                 : 14 s 720 ms
Bit rate mode                            : Variable
Bit rate                                 : 97.8 Mb/s
Maximum bit rate                         : 100.0 Mb/s
Width                                    : 3 840 pixels
Height                                   : 2 160 pixels
Display aspect ratio                     : 16:9
Frame rate                               : 25.000 FPS
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive
Bits/(Pixel*Frame)                       : 0.472
Stream size                              : 172 MiB (95%)
Color range                              : Limited
Color primaries                          : BT.709
Transfer characteristics                 : BT.709
Matrix coefficients                      : BT.709


  


• NumPy array of a video changes from the original after writing into the same video

  29 mars 2021, par Rashiq

  I have a video (test.mkv) that I have converted into a 4D NumPy array - (frame, height, width, color_channel). I have even managed to convert that array back into the same video (test_2.mkv) without altering anything. However, after reading this new, test_2.mkv, back into a new NumPy array, the array of the first video is different from the second video's array i.e. their hashes don't match and the numpy.array_equal() function returns false. I have tried using both python-ffmpeg and scikit-video but cannot get the arrays to match.

  


  Python-ffmpeg attempt :

  


  import ffmpeg
import numpy as np
import hashlib

file_name = 'test.mkv'

# Get video dimensions and framerate
probe = ffmpeg.probe(file_name)
video_stream = next((stream for stream in probe['streams'] if stream['codec_type'] == 'video'), None)
width = int(video_stream['width'])
height = int(video_stream['height'])
frame_rate = video_stream['avg_frame_rate']

# Read video into buffer
out, error = (
    ffmpeg
        .input(file_name, threads=120)
        .output("pipe:", format='rawvideo', pix_fmt='rgb24')
        .run(capture_stdout=True)
)

# Convert video buffer to array
video = (
    np
        .frombuffer(out, np.uint8)
        .reshape([-1, height, width, 3])
)

# Convert array to buffer
video_buffer = (
    np.ndarray
        .flatten(video)
        .tobytes()
)

# Write buffer back into a video
process = (
    ffmpeg
        .input('pipe:', format='rawvideo', s='{}x{}'.format(width, height))
        .output("test_2.mkv", r=frame_rate)
        .overwrite_output()
        .run_async(pipe_stdin=True)
)
process.communicate(input=video_buffer)

# Read the newly written video
out_2, error = (
    ffmpeg
        .input("test_2.mkv", threads=40)
        .output("pipe:", format='rawvideo', pix_fmt='rgb24')
        .run(capture_stdout=True)
)

# Convert new video into array
video_2 = (
    np
        .frombuffer(out_2, np.uint8)
        .reshape([-1, height, width, 3])
)

# Video dimesions change
print(f'{video.shape} vs {video_2.shape}') # (844, 1080, 608, 3) vs (2025, 1080, 608, 3)
print(f'{np.array_equal(video, video_2)}') # False

# Hashes don't match
print(hashlib.sha256(bytes(video_2)).digest()) # b'\x88\x00\xc8\x0ed\x84!\x01\x9e\x08 \xd0U\x9a(\x02\x0b-\xeeA\xecU\xf7\xad0xa\x9e\\\xbck\xc3'
print(hashlib.sha256(bytes(video)).digest()) # b'\x9d\xc1\x07xh\x1b\x04I\xed\x906\xe57\xba\xf3\xf1k\x08\xfa\xf1\xfaM\x9a\xcf\xa9\t8\xf0\xc9\t\xa9\xb7'


  


  Scikit-video attempt :

  


  import skvideo.io as sk
import numpy as np

video_data = sk.vread('test.mkv')

sk.vwrite('test_2_ski.mkv', video_data)

video_data_2 = sk.vread('test_2_ski.mkv')

# Dimensions match but...
print(video_data.shape) # (844, 1080, 608, 3)
print(video_data_2.shape) # (844, 1080, 608, 3)

# ...array elements don't
print(np.array_equal(video_data, video_data_2)) # False

# Hashes don't match either
print(hashlib.sha256(bytes(video_2)).digest()) # b'\x8b?]\x8epD:\xd9B\x14\xc7\xba\xect\x15G\xfaRP\xde\xad&EC\x15\xc3\x07\n{a[\x80'
print(hashlib.sha256(bytes(video)).digest()) # b'\x9d\xc1\x07xh\x1b\x04I\xed\x906\xe57\xba\xf3\xf1k\x08\xfa\xf1\xfaM\x9a\xcf\xa9\t8\xf0\xc9\t\xa9\xb7'


  


  I don't understand where I'm going wrong and both the respective documentations do not highlight how to do this particular task. Any help is appreciated. Thank you.

  


• ffmpeg Hardware acceleration (hw accel) has poor performance on Hololens 1

  12 avril 2021, par Dimitris

  I am building a library in C++ (dll) for decoding video h264 format. The problem is that on CPU is faster than GPU (hw accel).

  


  In the CPU, the execution time is about 1200 micro secs and the total time on dll is 2600 micro.

  


  In the GPU, the execution time is less, about 200 micro secs but transfer from GPU to CPU (RAM) has extra time, about 1100-1300 micro secs and the total time is 4200 micro. I am using the d3d11va/cuda hw accel device. That's all is not a problem on a PC, but I need the dll for the Hololens app (with poor performance) and I need to decode video on 30 fps in 1080p.

  


  My code is based on this tutorial from ffmpeg github repository.

  


  Do you have any suggestions on how to avoid the command av_hwframe_transfer_data(sw_frame, frame, 0) cause the bottleneck in the Hololens is about the transfer from GPU to CPU ?

  


  Thank you