Scikit-video allows users easy access to video files through the use of the FFmpeg/LibAV backend. This toolkit privides both high-level and low-level abstractions for reading and writing video files.
Scikit-video comes bundled with state-of-the-art quality measurement tools allowing users to begin curating their own video collections. Using the latest quality-aware tools allows researchers to easily compare their algorithms against a consistent and peer-reviewed set of tools.
Finally, scikit-video provides helpful utilities, like scene-boundary detectors and block-motion estimators commonly used in video processing algorithms.
You can use the function skvideo.setFFmpegPath or skvideo.setLibAVPath in the core skvideo API.
You may be using an old version of scikit-video. Uninstall it using
$ sudo pip uninstall sk-video
$ sudo pip install scikit-video
You can verify the module path by importing skvideo and printing the path to the init file
import skvideo
print(skvideo.__file__)
This should produce output like
/usr/lib/python*/site-packages/scikit_video-*.*.*-py*.egg/skvideo/__init__.pyc
You may not have supplied num_frames. The autodetection process sometimes requires two passes on an input video (depending on the type of video), which can make a huge difference for a large video corpus. By supplying num_frames, your code may speed up tremendously.
Please give us a full report of your problem in the issue tracker over on github.
Please fork our project, make a new branch on your fork, then make changes there. Improvements are always welcome, and we’d love to see an included benchmark and test to show it working nicely. If you want more complete instructions, check out the contributing doc:
https://github.com/scikit-video/scikit-video/blob/master/CONTRIBUTING.rst.