Skip to content

Last dimension of the npy feature map data is different from csv files #3

Description

@MusenZR

As discribed in the paper, the last dimension of feature map is suppose to be reflection intensity, the csv and *.mat files indicates that the fifth d data is a int number. But after loading the numpy feature map, it can be seen that the last d is a negative float number.I parsed the csv data into pytorch dataloader and run it in pytorch env, MSE loss show a difference from running the zipped numpy feature map.

from the numpy feature map
[[-0.45508 0.96094 -1.7871 0.71201 -0.53580797]
[-0.17871 1.6396 0.95117 0.356 -0.61807922]
[ 0. 2.0469 -0.61914 0.71201 -0.55637578]
[ 0.12988 2.0742 -0.055664 1.068 -0.53580797]
[ 0.18652 1.9561 0.32324 -0.356 -0.28899424]
[ 0.18945 1.9736 0.39355 0.71201 -0.26842643]
[ 0.18945 1.9883 0.30664 0.356 0.80109974]
[ 0.18945 1.9961 0.25195 -0.356 -0.12445175]]
same data from csv files:
[[-4.5508e-01 9.6094e-01 -1.7871e+00 7.1201e-01 9.0000e+00]
[-1.7871e-01 1.6396e+00 9.5117e-01 3.5600e-01 5.0000e+00]
[ 0.0000e+00 2.0469e+00 -6.1914e-01 7.1201e-01 8.0000e+00]
[ 1.2988e-01 2.0742e+00 -5.5664e-02 1.0680e+00 9.0000e+00]
[ 1.8652e-01 1.9561e+00 3.2324e-01 -3.5600e-01 2.1000e+01]
[ 1.8945e-01 1.9736e+00 3.9355e-01 7.1201e-01 2.2000e+01]
[ 1.8945e-01 1.9883e+00 3.0664e-01 3.5600e-01 7.4000e+01]
[ 1.8945e-01 1.9961e+00 2.5195e-01 -3.5600e-01 2.9000e+01]]

You can see the difference in the last dim.
So what's the last dim of the numpy feature map? Did you do some calculation to optimize the raw intensity?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

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