Skip to content

questions regarding multiplicative constant for flow & correlation max_displacement in multi-levels #7

Open
@InnovArul

Description

@InnovArul

I am trying to understand the code and the design decisions made while designing this model.
In this regard, I have two doubts as below:

  1. The correlation max_displacement is kept constant at 4 for all the levels of feature maps.
    self.corr = Correlation(pad_size=4, kernel_size=1, max_displacement=4, stride1=1, stride2=1, corr_multiply=1)

In your opinion, do you think we have to have different max_displacement for different levels of the network?

  1. I would like to get an intuition regarding the multiplication factors used for flow:
    0.625 for level-6 flow
    feat5b_warped = self.warp(feat5b, 0.625*flow6_fwd_up)
    feat5c_warped = self.warp(feat5c, -0.625*flow6_bwd_up)

1.25 for level-5 flow

feat4b_warped = self.warp(feat4b, 1.25*flow5_fwd_up)
feat4c_warped = self.warp(feat4c, -1.25*flow5_bwd_up)

2.5 for level-4 flow

feat3b_warped = self.warp(feat3b, 2.5*flow4_fwd_up)
feat3c_warped = self.warp(feat3c, -2.5*flow4_bwd_up)

5 for level-3 flow

feat2b_warped = self.warp(feat2b, 5.0*flow3_fwd_up)
feat2c_warped = self.warp(feat2c, -5.0*flow3_bwd_up)

Could you please let me know why do we need these multiplicative factors?
I am sorry if this is a basic question to be asked. As the network is learnable, isn't it possible that the network learns this multiplicative factor automatically as well?

Thanks in advance!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions