yuehaowang / recolornerf Goto Github PK

Thanks for your great work!
I'm interested in the reference time for one image. If I change the color in Real-time, how much should the maximum resolution be?

Hello, when I try to generate a new palette, I find that I need poses_bounds.npy' document.
What kind of file is this? There is no such file in the official release of LLFF dataset.
Have you generated his script file?

Thank you for your good work！
I would like to recolor a local area of the target, but from what the paper says, it seems that this would change the overall color.

Hi,
When I run the color_edit.py, I met the error below.

  File "/rockywin.wang/NeRF/RecolorNeRF/color_edit.py", line 141, in <module>
    display(box_auto)
NameError: name 'display' is not defined

I copyed the color_edit.ipynb to color_edit.py.

Hi,
Thanks for the great work!
I am trying to run your code but reported an import error as shown in the title. Also, I did not find the requested module in "Supported datasets" of readme file.
The traceback location is shown as follow:

models/palette_tensoRF.py
class PLTRender(torch.nn.Module):
def init():
self.n_dim = 3 + len_palette
......
layer1 = torch.nn.Linear(self.in_mlpC, featureC)
layer2 = torch.nn.Linear(featureC, featureC)
layer3 = torch.nn.Linear(featureC, len_palette - 1)
torch.nn.init.constant_(layer3.bias, 0)
self.mlp = torch.nn.Sequential(
layer1, torch.nn.LeakyReLU(inplace=True),
layer2, torch.nn.LeakyReLU(inplace=True),
layer3)
self.n_dim += 1
I recently tried to integrate this work of yours on instant-ngp.
But I do not understand the role of self.n_dim, and why the output dimension of layer3 is len_palette - 1.
Besides，why is the activation function after each layer LeakyReLU, and the Relu function is also used in TensoRF.

Hope you can help me

Best wish！

You said in your paper :
“we aim to penalize the L0 norm of these components, i.e., k [𝑊1 , . . . , 𝑊𝐾 ] k 0 . Since the L0 norm is not differ-
entiable, we design a specialized soft counting norm”.
But the L1 norm is the optimal convex approximation of the L0 norm, and it is easier to optimize and solve than the L0 norm, so why don't you use the L1 norm to achieve regularization.

Hi friend, thanks for your previous answer.
When I run color_edit.ipynb, there is no GUI that I would expect to allow me to easily edit colors.
What parameters should I edit to edit the nerf color?
Or just edit the order of color layers in get_palette.ipynb or add new color layers？

Hi, thank you for your work!
Please, why are the colors given for the palette parameters different from the palette colors in the article such as fern?

I tried fitting the color and density with an MLP like the original nerf. There is no problem with training and testing, but when doing color editing, there is a problem.

Test images during training：

Result after color editing：