VIDI

Python wrapper to FFMPEG

v.0.2 wipe simplified and rewritten

• Linux only (for windows replace 'ffmpeg' with 'ffmpeg.exe' in vidi.ff.py - untested)

Exposes a subset of ffmpeg functionality wrapped in functions. VIDI's main intention is interaction between video and array data, numpy and torch.

Converts RGB to and from FourCC and higher bits formats 10,12,14 and 16 bit files.

---

classes

vidi.FF # requires numpy, cv2

V = FF(videofile)
V.export_clip()
V.export_frames()
V.to_numpy()        # -> ndarray
V.make_subtitles()  # frame number and time subtitles
V.vlc()             # play with vlc
V.play()            # ffplay
V.stats             # ffprobe -> dict

vidi.FFDataset

with FFDataset(videofile) as D:
    D.__getitem__() # -> torch.tensor if torch installed else np.ndarray
    D.__len__()

standalone functions

make_subtitles(videofile)   # creates subtitle file with frame numbers
export_frames(videofile)    # exports frame range to images
export_clip(videofile)
get_formats()               # IO vidi subset of installed ffmpeg supported video formats
images_to_video()           # creates video from images
yxx2rgb()                   # FourCC (YCC and YUV) to RGB
rgb2yxx()                   # RGB to FourCC (YCC and YUV)
yxx_matrix()
expand_fourcc() 4nm -> 444  # uncompress raw FourCC flat arrays to 3 channels
# todo expose compress_fourcc()

TODO:

• unit tests;
• multiprocessing to torch dataset