How to get the output as your project page

[Dandelionym]

Excellent work! One question I want to ask:

I got the shape by demo.py like this:

But your paper gives the result like this:

Question: How to get the second one?

[pengwangucla]

+1 how to recover the mesh after having these point cloud ? Or the demo output is just a group of enlarged point clouds

Or did you directly learn a nerf model from the point cloud for visualizing these results. If that is the case, wonder when the code could be released ?

[Dandelionym]

@chaoyuaw @Celebio Hope the answer can be replied to soon. :-)

[gkioxari]

Hi all! The output of MCC is a 3D point cloud. Namely it's a set of colored 3D points: { $(x,y,z), (r,g,b)$ }. The visualizations we provide in the project page and the paper are Blender renderings of those colored 3D point clouds. There is no mesh extraction stage nor did we learn a NeRF model!

We also provide plotly visualizations in embedded in html (click on the 3D button in the project page). Alternatively, you can use PyTorch3D to render 3D point clouds to get 2D images or any other render you prefer!

[Dandelionym]

Thank you @gkioxari, you mean the results in your project page are from Blender using the output of MCC, which is the 3d points and colors.

I will try and if possible, some question may seek for your help. >:-)

[gkioxari]

I realize that you might not be familiar with how to visualize 3D data. For completeness here is an explanation. The output of MCC is a 3D colored point cloud. To visualize any such 3D point cloud, one must use a renderer. In this particular paper, we use Blender to visualize the MCC outputs. In our webpage, we also visualize the results with a plotly which is another visualization scheme by rendering and which is interactive (click the 3D button).

[Dandelionym]

Thank you, professor @gkioxari, I've tried that and successfully done it. Thank you again.