search

huangjh-pub / neural-galerkin

[SIGGRAPH Asia 2022] A Neural Galerkin Solver for Accurate Surface Reconstruction
MIT License
94 stars 8 forks source link

how to test own pointcloud ? #2

Open zuixiaosanlang opened 1 year ago

zuixiaosanlang commented 1 year ago

Thanks for your great work!

1）what is the steps for data preprocessing?

i normalize to unit cube like this: LOD_input, normals = load_ply(ply_path)

LOD_input_min = np.min(LOD_input, 0) LOD_input_max = np.max(LOD_input, 0) LOD_input_scale = np.max(LOD_input_max - LOD_input_min)

LOD_input = LOD_input - np.reshape(LOD_input_min, [1, 3]) LOD_input = LOD_input / LOD_input_scale input_points = LOD_input - np.reshape([0.5, 0.5, 0.5], [1, 3])

I don't know if that's true

2）the result i get seem not very good. vis the SubsamplePointcloud (N=10000): image

get the mesh image there is no points but output mesh

image mesh seems not very good

Am I missing something？？ Any suggestions？？

heiwang1997 commented 1 year ago

Hi @zuixiaosanlang sorry for the late response.

For 1), I think this should be correct if your data is already y-up. For 2) this seems weird. May I know which model you're using? If you want you may also give me a link of your test point cloud so that I can try it on my end.

zuixiaosanlang commented 1 year ago

@heiwang1997 i am sorry to reply so late. my input data: test_data