Currently, most flow field functions assume that the number of vector field channels is equal to the number of grid dimensions. However, in case of a spatio-temporal flow field, each vector field only deforms space.
The shape of a spatio-temporal flow field tensor is (N, D, T, ..., X), i.e.,
(N, 2, T, Y, X) for a 2D+t flow field, and(N, 3, T, Z, Y, X) for a 3D+t flow field.
Also the Grid would only apply to the spatial dimensions. The spacing between time points along the temporal dimension may need to be stored in a separate metadata field.
At the moment, one may need to use a sequence / batch of 2D or 3D flow fields instead, but computing a temporal partial derivative is then also not supported by the respective image functions.
Currently, most flow field functions assume that the number of vector field channels is equal to the number of grid dimensions. However, in case of a spatio-temporal flow field, each vector field only deforms space.
The shape of a spatio-temporal flow field tensor is
(N, D, T, ..., X), i.e.,(N, 2, T, Y, X)for a 2D+t flow field, and(N, 3, T, Z, Y, X)for a 3D+t flow field.Also the
Gridwould only apply to the spatial dimensions. The spacing between time points along the temporal dimension may need to be stored in a separate metadata field.At the moment, one may need to use a sequence / batch of 2D or 3D flow fields instead, but computing a temporal partial derivative is then also not supported by the respective image functions.