You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We do have colorview/channelview/rawview to build these functionalities, but from a user's perspective, it would be much much easier to understand and use things like:
to_matlab/from_matlab
to_torch/from_torch
Because other framework doesn't always contain colorspace information on their numerical array, assumptions must be made. For instance:
for MATLAB we can assume that 3D numerical array with size (m, n, 3) is an RGB image of size (m, n).
for deep learning frameworks like pytorch, we can assume that 4D tensor with size (w, h, 3) is an RGB image of size (m, n).
These utilities do not need to cover every use case and do not need to provide 100% consistency; they're meant to help users to migrate the old codes from other frameworks.
The text was updated successfully, but these errors were encountered:
We do have
colorview
/channelview
/rawview
to build these functionalities, but from a user's perspective, it would be much much easier to understand and use things like:to_matlab
/from_matlab
to_torch
/from_torch
Because other framework doesn't always contain colorspace information on their numerical array, assumptions must be made. For instance:
(m, n, 3)
is an RGB image of size(m, n)
.(w, h, 3)
is an RGB image of size(m, n)
.These utilities do not need to cover every use case and do not need to provide 100% consistency; they're meant to help users to migrate the old codes from other frameworks.
The text was updated successfully, but these errors were encountered: