zhuker/pyvideohacks

Load video into ML pipeline

Load videos into your ML pipelines

loader = SimpleVideoLoader("testdata/bipbop15_270_stereo.mp4")
for frame in loader:
    print(frame.shape)

Output

(272, 480, 3)
(272, 480, 3)
...

Load HLS

Read first 5 frames from public HLS url

loader = SimpleVideoLoader("https://test-streams.mux.dev/x36xhzz/x36xhzz.m3u8")
for frame in itertools.islice(loader, 5):
    print("frame from live video", frame.shape)

Output

...
frame from live video (720, 1280, 3)
frame from live video (720, 1280, 3)
...

Force FPS

The following will force fps to 5, ffmpeg will drop frames to achieve the fps

loader = SimpleVideoLoader("testdata/bipbop15_270_stereo.mp4", forcefps=5)
count = 0
for frame in loader:
    print(frame.shape)
    count += 1

print("read", count, "frames")
assert count == 77