Has anyone successfully segmented the pointcloud generated by Gaussian Splatting?
I have a studio-like environment with 20+ cameras 360 deg around a person and I used Gaussian Splatting to recon the scene. RGB looks great. But when I tried extracting the pointcloud from splatts positions there are some noisy points around the person. My goal is to get the clean pointcloud of the person.
Conventional clustering method seems do not work well in this case. I tried K-means and DBSCAN from sklearn and open3d and the noisy points are still not filtered out. The only one works is the label connected components method from CloudCompare. However, there's no efficient label connected components implementation in python for pointcloud yet.
Has anyone else faced similar issues and if so, appreciate any pointers to resolve this issue!
Has anyone successfully segmented the pointcloud generated by Gaussian Splatting?
I have a studio-like environment with 20+ cameras 360 deg around a person and I used Gaussian Splatting to recon the scene. RGB looks great. But when I tried extracting the pointcloud from splatts positions there are some noisy points around the person. My goal is to get the clean pointcloud of the person.
Conventional clustering method seems do not work well in this case. I tried K-means and DBSCAN from sklearn and open3d and the noisy points are still not filtered out. The only one works is the label connected components method from CloudCompare. However, there's no efficient label connected components implementation in python for pointcloud yet.
Has anyone else faced similar issues and if so, appreciate any pointers to resolve this issue!