add RaDeGS #317
add RaDeGS #317
Conversation
|
Hello, how to test for creating mesh from existing PLY with this PR? |
| @@ -400,7 +423,7 @@ def rasterization( | |||
| "tile_size": tile_size, | |||
| "n_cameras": C, | |||
| } | |||
| return render_colors, render_alphas, meta | |||
| return render_colors, render_alphas, expected_depth, median_depths, expected_normal, meta | |||
There was a problem hiding this comment.
If this is meant to be merged, tests and the other training scripts in examples need to be adapted to this API change.
| image = imageio.imread(self.parser.image_paths[index])[..., :3] | ||
| image = imageio.imread(self.parser.image_paths[index]) | ||
| if image.shape[-1]==4: | ||
| mask = image[...,-1] |
There was a problem hiding this comment.
Should the conversion to a binary mask be done here? (you're doing data["mask"]/255<0.5 later on)
|
Any news on this PR? |
| mat3x2<T> nJ_inv_T = mat3x2<T>(vv + 1, -uv, // 1st column | ||
| -uv, uu + 1, // 2nd column | ||
| -u, -v // 3nd column | ||
| ); | ||
| T factor = l / (uu + vv + 1); | ||
| vec3<T> Cinv_uvh_n = glm::normalize(Cinv_uvh); | ||
| T u_Cinv_u_n_clmap = max(glm::dot(Cinv_uvh_n, uvh), 1E-7); |
There was a problem hiding this comment.
Hello @BaowenZ , I am a little bit lost here and can not align the code with the equation in original paper. Is there any hint?
There was a problem hiding this comment.
This is related to further derivation of the rendering equation. The plane vector is
|
Hi, Ruilong is currently writing the code with a similar depth rendering method. And this PR might be helpful as a reference. @ichsan2895 |
|
hi @BaowenZ can the output of RaDeGS be used in a typical gaussian splat renderer? Or does it require special modifications? |
|
Appreciate your amazing work! I have a question about the mesh extraction part of the simple_trainer_recon.py.
|
Hi! I added the Rade-GS depth and normal rendering.