chengkunli96 / kinectfusion Goto Github PK

Hi @chengkunli96, thanks for sharing this work. I have tried to test on TUM datasets, when use the rgbd_dataset_freiburg2_xyz, the pointcloud results look normal, but when I try to use rgbd_dataset_freiburg3_walking_xyz, I only get one person's part hand point cloud other than the whole image[the camera_config.txt have modified based on the freiburg3]

I tried to modify the ds (default=1)vaule, it seems the result points cloud can show more body part, but still can not show the whole image content.

what's more, when I use the fusion_by_point_cloud.py on the rgbd_dataset_freiburg3_walking_xyz, I always get indexError that index n is out of bouds for axis 0 with size n. it seems this scirpts

Hope you can give some advice, thanks~

[update]
I have solved the indexError, but the dynamic fusion point result are bad.

Hey for reproducability may also provide the requirements.txt file

so this is not an issue more a recommendation. thanks for the cool work!

numba==0.55.1
numpy==1.21.5
open3d==0.15.1
opencv-python==4.5.5.64
Pillow==9.0.1
pycparser==2.21
pyglet==1.5.22
Pygments==2.11.2
PyOpenGL==3.1.0
pyparsing==3.0.7
pyrender==0.1.45
scikit-image==0.19.2
scipy==1.8.0
trimesh==3.10.2