I would like to load a video file's frames into a numpy array. I want the frames to be properly upright, which means I need to read the orientation metadata in the video file, and rotate the loaded frames accordingly.
I have a means of loading the frames (opencv's python bindings), so all I need is a way to read the video file's orientation metadata.
I'm trying to read a .MOV file recorded off my iPhone, in python running on an Ubuntu 14.04 machine.
Stuff I've looked at:
- opencv: no way of inspecting video orientation
- hachoir-metadata: doesn't display orientation metadata
- enzyme: only works on .mkv files, AFAICT.
- ffmpeg: no longer available as a stand-alone executable on Ubuntu's repositories. This is a dealbreaker since I'd rather just hand-specify the rotation rather than go through the trouble of compiling ffmpeg from source just to get video orientation.
- PIL, pyexiv: AFAICT, only loads image files.
I'm running the code on an Ubuntu 14.04 machine, so I'm looking for libraries or command-line programs available on that platform.
Any pointers would be greatly appreciated.
