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
Users may like to downsample a signal consistently, which is most straightforward in harmonic space. This may involve starting from a high resolution pixel-space map, computing harmonic coefficients and then inverse transforming at a lower resolution. If this is the goal, we may instead be able to terminate the recursive steps early and return a downsampled map with only the overhead of a forward and inverse pass at the lower resolution.
Another use case for this would be for convolutional layers in a UNet type network. These layers fall into the same class.
The text was updated successfully, but these errors were encountered:
Users may like to downsample a signal consistently, which is most straightforward in harmonic space. This may involve starting from a high resolution pixel-space map, computing harmonic coefficients and then inverse transforming at a lower resolution. If this is the goal, we may instead be able to terminate the recursive steps early and return a downsampled map with only the overhead of a forward and inverse pass at the lower resolution.
Another use case for this would be for convolutional layers in a UNet type network. These layers fall into the same class.
The text was updated successfully, but these errors were encountered: