Video object returns array of zeros on second call

[EricThomson]

Thanks a lot for the useful package! Sorry if my question betrays a basic misunderstanding:

I have an avi file that I am reading and pims works great accessing individual frames, initially:

pims_movie = pims.Video('filename.avi')
for ind in range(10,12):
    movie_frame = pims_movie[ind]
    print("\n", ind)
    print(movie_frame[:5, :5, 0])

It prints out reasonable values:

10
[[ 85 87 89 90 93]
[ 87 89 90 92 96]
[ 88 89 91 93 97]
[ 89 92 93 94 97]
[ 92 94 96 98 102]]

11
[[ 86 88 90 90 91]
[ 90 89 90 93 95]
[ 90 90 91 95 96]
[ 91 93 95 96 98]
[ 94 96 97 99 102]]

But if I run the for loop again (without reloading the movie), I get arrays of zeros back:

10
[[0 0 0 0 0]
[0 0 0 0 0]
[0 0 0 0 0]
[0 0 0 0 0]
[0 0 0 0 0]]

11
[[0 0 0 0 0]
[0 0 0 0 0]
[0 0 0 0 0]
[0 0 0 0 0]
[0 0 0 0 0]]

Is this expected behavior? Am I supposed to be closing out the video and reloading it, or using a context manager, or something?

[pgunn]

It looks like there are a few more ways to confuse the pims handle:

>>> pimfh = pims.Video('/mnt/home/pgunn/caiman_data/example_movies/msCam13.avi')
>>> pimfh[100][100].max()
128
>>> pimfh[100][100].max()
128
>>> data = np.array(pimfh)[200:250,100:120,500:510,0]
>>> pimfh[100][100].max()
0

[pgunn]

The code I pasted above shows the issue with pims 0.6.1

[pgunn]

In case it's helpful, the data file I'm using is available here (it's a dataset used as part of a scientific software package):

https://caiman.flatironinstitute.org/~neuro/caiman_downloadables/msCam13.avi

[pgunn]

I found a very weird way to reset the pims handle back to a healthy state, which may hint as to what's going on:

>>> pimfh[100][100].max()
128
>>> data = np.array(pimfh)[200:250, 100:120, 500:510, 0]
>>> pimfh[100][100].max()
0
>>> _ = pimfh[1]
>>> pimfh[100][100].max()
0
>>> _ = pimfh[0]
>>> pimfh[100][100].max()
128

I'll still rely on the pims developers to fully understand (and ideally patch) it, but we may be able to workaround the bug for now.

[pgunn]

In digging into this, it's actually not pims' fault, as we've found a way to reproduce this using moviepy calls (which is the backend both of us are using).

>>> mph = moviepy.editor.VideoFileClip(fn)
>>> mph.size
[752, 480]
>>> mph.duration
50.0
>>> mph.fps
20.0
>>> np.array(mph.get_frame(0)).max()
173
>>> np.array(mph.get_frame(2)).max()
206
>>> np.array(mph.get_frame(1)).max()
0
>>> np.array(mph.get_frame(0)).max()
173
>>> np.array(mph.get_frame(1)).max()
188

We'll reach out to the moviepy project.

[pgunn]

Zulko/moviepy#2044