skvideo.io.vreader always capturing the same, fixed frame about scikit-video HOT 7 OPEN

That's quite strange. Seems like another mobile-video issue. What happens when you just ask ffmpeg to extract the frames? Command to do this:
ffmpeg -i my_video.webm %d.jpg

from scikit-video.

Two things happen:

1. ffmpeg extracts 2881 frames out of a 3 second video. this must be a ffmpeg bug
2. frames are not identical

On the other hand, in other cases where the number of frames extracted by ffmpeg matches the number of frames extracted by skvideo.io.vread, (and when the number of frames is reasonable given the video duration) you can clearly infer motion from the skvideo.io frames.

It appears that, in presence of this ffmpeg bug, skvideo.io does not know what to do and always extracts the same frame.

Also, I know it may be too much to ask, but would this be a possible workaround (not really a fix) when this happens: infer a frames per second rate and extract the frames accordingly in skvideo.io.vread?

Example:: this video is a 3 sec video and has 2881 frames. Assuming 20 frames per second, there should be 60 frames. 2281 divided by 60 is 48, could skvideo.io extract every 48th frame (frames 48, 96, 144, 192, etc)?

Let me know if you would like me to email a sample video (the one where ffmpeg extracts 2881 frames in 3 secs).

from scikit-video.

Please do email a sample video. I'm sure there should be enough metadata in the video file that we just aren't parsing in scikit-video. As for the ffmpeg issue, I guess it's fetching an invalid framerate?

from scikit-video.

A sample program for reading,playing and writing the frames using skvideo and opencv

video = 'car.avi'
videogen = skvideo.io.vread(video)
videometadata = skvideo.io.ffprobe(video)
frame_rate = videometadata['video']['@avg_frame_rate']
num_frames = np.int(videometadata['video']['@nb_frames'])
print num_frames
for i in range(0,100): % I restricted to 100 as an example but if you wish to write all the frames of video change to for i in range(0,num_frames)
    frame = cv2.cvtColor(videogen[i],cv2.COLOR_BGR2RGB) %this code do color conversion
    cv2.imwrite("frame_"+str(i)+".jpg",frame)
    cv2.imshow('frame',videogen[i])
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
cv2.destroyAllWindows()

Hope this might help someone.

from scikit-video.

Hi there. I see the issue has been closed but without a solution.
We are experiencing the same problem. Here is the case: we have a video recorded on an iPhone at a nominal rate of 120 FPS. The iPhone doesn't really keep up with it, so it records the video at 118 FPS.

Here is the output from ffprobe:

ffprobe version 3.4 Copyright (c) 2007-2017 the FFmpeg developers
  built with gcc 7.2.0 (crosstool-NG 8a21ab48)
  configuration: --prefix=/home/chiara/anaconda3/envs/mocap --disable-doc --enable-shared --extra-cflags='-fPIC -I/home/chiara/anaconda3/envs/mocap/include' --extra-cxxflags='=-fPIC' --extra-libs='-L/home/chiara/anaconda3/envs/mocap/lib -lz' --enable-pic --disable-static --disable-gpl --disable-nonfree --disable-openssl --enable-libvpx --cc=/home/chiara/anaconda3/envs/mocap/bin/x86_64-conda_cos6-linux-gnu-cc --cxx=/home/chiara/anaconda3/envs/mocap/bin/x86_64-conda_cos6-linux-gnu-c++ --enable-libopus
  libavutil      55. 78.100 / 55. 78.100
  libavcodec     57.107.100 / 57.107.100
  libavformat    57. 83.100 / 57. 83.100
  libavdevice    57. 10.100 / 57. 10.100
  libavfilter     6.107.100 /  6.107.100
  libswscale      4.  8.100 /  4.  8.100
  libswresample   2.  9.100 /  2.  9.100
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'IMG_1967.MOV':
  Metadata:
    major_brand     : qt  
    minor_version   : 0
    compatible_brands: qt  
    creation_time   : 2018-05-08T04:29:42.000000Z
    com.apple.quicktime.description: RAD:1506.666382:1506.666382:967.640320:515.699402:120.000000
  Duration: 00:00:10.57, start: 0.000000, bitrate: 23360 kb/s
    Stream #0:0(und): Video: hevc (Main) (hvc1 / 0x31637668), yuvj420p(pc, progressive), 1920x1080, 22861 kb/s, 117.63 fps, 600 tbr, 600 tbn, 600 tbc (default)
    Metadata:
      creation_time   : 2018-05-08T04:29:42.000000Z
      handler_name    : Core Media Data Handler
      encoder         : 'hvc1'

The problem that we experience is that the reported FPS is 600. Now, if we iterate through the frames, the iterator returns only the correct number of frames, but actually it returns 5 copies of the same frame and then stops at a fifth of the video.

I cannot upload the video here, but I can email it to whoever needs to handle this issue.

@beyondmetis can you reopen this, or would you like me to open a new issue?

Thanks

from scikit-video.

Let’s solve this. Please send the video to my email, [email protected]

from scikit-video.

Has this been solved? I still see the same problem with version scikit-video==1.1.11

from scikit-video.