About suface_normal Prediction

[amnesiac555]

Hello, JiaxiongQ!
when i read your code, i found that after you predict the surface normal vector, you perform some transformation operations on it. the code is:
outputN = torch.zeros_like(normals2)
outputN[:,:,:,0] = -normals2[:,:,:,0]
outputN[:,:,:,1] = -normals2[:,:,:,2]
outputN[:,:,:,2] = -normals2[:,:,:,1]
I'm confused about this part. Can you explain it，Thank you very much.

[JiaxiongQ]

When estimating the ground-truth of surface normal from depth, we did the same operation to generate the color we want. So the surface normal from predicted depth should do this operation again.

[JennyGao00]

Can you explain your method to estimate the ground-truth of surface normal from depth?
Thanks a lot.

[JiaxiongQ]

For KITTI, we fitted a plane for sparse depth points in a window. Please vist https://github.com/Cindy-xdZhang/surface-normal for details.

…

On Wed, Jan 13, 2021 at 10:07 PM JennyGao00 ***@***.***> wrote: Can you explain your method to estimate the ground-truth of surface normal from depth? Thanks a lot. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#18 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AJANJRCJHBDP7CHUBRJVV5LSZWSLLANCNFSM4JTJP4GA> .

[JennyGao00]

For KITTI, we fitted a plane for sparse depth points in a window. Please vist https://github.com/Cindy-xdZhang/surface-normal for details.
…
On Wed, Jan 13, 2021 at 10:07 PM JennyGao00 @.***> wrote: Can you explain your method to estimate the ground-truth of surface normal from depth? Thanks a lot. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#18 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJANJRCJHBDP7CHUBRJVV5LSZWSLLANCNFSM4JTJP4GA .

thanks for your reply!
I have ran the above code, but i wonder, numerically, what should f, fx, fy of the nuydepthv2 dataset be? (like 1, 1, 1, or some other numbers?)

[JiaxiongQ]

You can find these parameters in their official toolbox. On Sun, Jan 24, 2021 at 11:26 PM JennyGao00 <notifications@github.com> wrote:

…

For KITTI, we fitted a plane for sparse depth points in a window. Please vist https://github.com/Cindy-xdZhang/surface-normal for details. … <#m_3864882966769329341_> On Wed, Jan 13, 2021 at 10:07 PM JennyGao00 *@*.***> wrote: Can you explain your method to estimate the ground-truth of surface normal from depth? Thanks a lot. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#18 (comment) <#18 (comment)>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJANJRCJHBDP7CHUBRJVV5LSZWSLLANCNFSM4JTJP4GA . thanks for your reply! I have ran the above code, but i wonder, numerically, what should f, fx, fy of the nuydepthv2 dataset be? (like 1, 1, 1, or some other numbers?) — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#18 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AJANJRAG3F2DXT7WQMXEDGDS3Q3ZBANCNFSM4JTJP4GA> .