robot0321

robot0321

In the Ph.D course in South Korea

Member Since 6 years ago

Experience Points
5
follower
Lessons Completed
1
follow
Lessons Completed
9
stars
Best Reply Awards
25
repos

30 contributions in the last year

Pinned
⚡ MATLAB simulation of visual-inertial odometry (VIO) & visual-wheel odometry
Activity
Oct
26
2 days ago
Activity icon
issue

robot0321 issue comment Totoro97/NeuS

robot0321
robot0321

Learning with 360-view data

Hi there,

I'm trying to learn 3D scene with 360-degree view dataset like lego data in NeRF. It seems that the learning process is suffered from the initial error of SDF estimation. Of course, it is a different scenario with IDR dataset, however, the learning usually converges to local optimum Could you please let me know is there any strategy to learn the 360 view data properly?

Oct
25
3 days ago
Activity icon
issue

robot0321 issue Totoro97/NeuS

robot0321
robot0321

Learning with 360-view data

Hi there,

I'm trying to learn 3D scene with 360-degree view dataset like lego data in NeRF. It seems that the learning process is suffered from the initial error of SDF estimation. Of course, it is a different scenario with IDR dataset, however, the learning usually converges to local optimum Could you please let me know is there any strategy to learn the 360 view data properly?

Activity icon
issue

robot0321 issue comment Totoro97/NeuS

robot0321
robot0321

Learning with 360-view data

Hi there,

I'm trying to learn 3D scene with 360-degree view dataset like lego data in NeRF. It seems that the learning process is suffered from the initial error of SDF estimation. Of course, it is a different scenario with IDR dataset, however, the learning usually converges to local optimum Could you please let me know is there any strategy to learn the 360 view data properly?

robot0321
robot0321

Thanks for providing the files! It really helps. The problem was the definition of camera coordinate. I totally forgot about that. image

The problem is solved by flipping the y,z axes. I really appreciate your generous help. Thank you very much! 👍 👍

Oct
24
4 days ago
Activity icon
issue

robot0321 issue comment Totoro97/NeuS

robot0321
robot0321

Learning with 360-view data

Hi there,

I'm trying to learn 3D scene with 360-degree view dataset like lego data in NeRF. It seems that the learning process is suffered from the initial error of SDF estimation. Of course, it is a different scenario with IDR dataset, however, the learning usually converges to local optimum Could you please let me know is there any strategy to learn the 360 view data properly?

robot0321
robot0321

@Totoro97 Thanks for your kind answer!

BTW, I suspect the problem is from data conversion. I just set the 'scale_mat' to np.eye(4) because the lego is almost placed in a unit cube ([-1,1]) I tried to learn the 'preprocess_cameras.py' code in DTU as you mentioned in README, but it does not work because they do not consider the 360-view scenario.

Now I got only a smoothed (not detailed) surface, so I guess there are errors in the scale_mat I also try to set the s=10 and 100 in SDF network, but they fail to estimate detailed surfaces Could I borrow any comment about this situation?

image

Oct
21
1 week ago
Activity icon
issue

robot0321 issue comment mikedh/trimesh

robot0321
robot0321

add texture from image to shape

Hi,

I have a 3D shape loaded from an .off file, and I have an image load as a PIL image.

I want to add this image as a texture to the mesh but I can't figure out how to do it. I did the following manipulation, but when I show the mesh, the texture is not applied.

import trimesh
from PIL import Image

im = Image.open("image.png")
m = trimesh.load("mesh.off", process=False)

tex = trimesh.visual.TextureVisuals(image=im)
m.visual.texture = tex

m.show()

How can I do this? Thank you in advance.

robot0321
robot0321

image

In my case, the .obj file contains uv information. Note that if the mesh.show() present a dark texture, use pyrender library

Oct
20
1 week ago
Activity icon
issue

robot0321 issue Totoro97/NeuS

robot0321
robot0321

Learning with 360-view data

Hi there,

I'm trying to learn 3D scene with 360-degree view dataset like lego data in NeRF. It seems that the learning process is suffered from the initial error of SDF estimation. Of course, it is a different scenario with IDR dataset, however, the learning usually converges to local optimum Could you please let me know is there any strategy to learn the 360 view data properly?

Oct
16
1 week ago
Activity icon
issue

robot0321 issue comment albertpumarola/D-NeRF

robot0321
robot0321

blender projects for the datasets

Hi, this is a really great work on dynamic scenes! The datasets you provided are really cool and I believe they are generated using blender, right? Can you share the blender project files you used to generate these datasets, as I am really interested to explore some other scenarios of dynamic scenes. Thank you very much!

robot0321
robot0321

if you check the blend files from NeRF dataset, you can control the scene using 'Controlpanel' by adjusting the value of 'Location Z'. I'm also a newbie in Blender, but I guess it is easy to generate other dynamic scenes.

Sep
23
1 month ago
Activity icon
issue

robot0321 issue sxyu/plenoctree

robot0321
robot0321

The name of "Plen"Octree

Hi there, thanks for your nice work.

I wonder about the meaning of "PlenOctree" instead of just Octree. please be kind if I miss any explanation.

Thanks!

Aug
27
2 months ago
Activity icon
fork

robot0321 forked yenchenlin/nerf-pytorch

⚡ A PyTorch implementation of NeRF (Neural Radiance Fields) that reproduces the results.
robot0321 MIT License Updated
fork time in 2 months ago
Aug
26
2 months ago
Aug
25
2 months ago