Open hellokenlee opened 8 years ago
I made a minimal example Here. Hope it helps.
Hi! I have a question about the 2-D matrix V. In the example you provided, the values of V are the same for all pixels. But sometimes, for different pixels, the neighborhood cost should be different, and I think that each pixel in an image should be corresponding to a V. are the values of V fixed for all pixels in this module?
Hi @zhangwenyuan0926,
are the values of V fixed for all pixels in this module?
Unfortunately yes, they are fixed. I wrote the fastmin
module a long time ago and I didn't keep it updated with the rest of the API. The good news is that the C++ implementations for aexpansion_grid_step
and abswap_grid_step
are unnecessary now, and they can be easily implemented in Python with the new API (or so I want to believe). I am planning to update this in the future and get rid of the old C++ fastmin
implementation.
In the meantime, you can just reuse the implementation of aexpansion_grid
(or abswap_grid
) from fastmin.py
and write your own aexpansion_grid_step
(or abswap_grid_step
) that builds the graph using the neighborhood costs you want.
Sorry for the inconvenience. I will fix this as soon as I have time.
Thank you so much for writing this package! The documentation is great in the maxflow part. But it's a little confusing in the maxflow.fastmin part... It would be great if you can improve the documentation on this part. Thank you again.