Skip to content

Temporal Loss - code vs. paper #6

@OriLifschitz

Description

@OriLifschitz

Hello, impressive work and masterfully written paper!

This is the code for the temporal loss:
`

temporal loss

temporal_loss = 0
Laplacian = torch.squeeze(C_x + C_y)
for i in range(1, time_frame - 1):
temporal_loss += torch.sum(mask_cur[i:i + 1, 0, :, :] * (
(height_map_cuda[i + 1, :, :] + height_map_cuda[i - 1, :, :] - 2 * height_map_cuda[i, :, :] - 0.5 * Laplacian[i, :, :]) ** 2)
)
`
While this is the equation taken from the paper:
image

I am a bit confused regarding where did the constant 'c' in eq. (13) ("c is the magnitude of the velocity") disappeared in the code please?
Or where in the paper can I read about this change between paper and code please?

Thanks a lot !

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions