[IPOL discuss] various questions about python demos

[eml]

> My problem with this is the input. Currently the input system is
> tailored to images, so it will take some coding to adapt it to another
> kind of input, for instance to  upload a text file, rename it, etc.
>
> Have you guys tried to do something similar?

Yes, our demos for optical flow take as (optional) input image, a
vector field which is stored in a 2-channel TIFF image.  These images
are not supported by the current demo system, because the PIL library
can not read them.  The amount of coding that you have to do to
overcome this problem is minimal: you simply override the function
"input_upload" to do what you need to do(instead of using the default
base_app.input_upload).  The only problematic part is the call to
"process_input", which you need to adapt from converting images using
PIL to read your graphs.