How to change the number of 3d Gauss initializations about gauhuman HOT 5 CLOSED

Hi, thanks for your interest in our work. You can specify the number of 3D Gaussians for the refined ZJU_MoCap dataset at

from gauhuman.

Hi, thanks for your interest in our work. You can specify the number of 3D Gaussians for the refined ZJU_MoCap dataset at

GauHuman/scene/dataset_readers.py

Line 726 in 732ec0d

pcd = BasicPointCloud(points=xyz, colors=SH2RGB(shs), normals=np.zeros((num_pts, 3)))

.

Thanks for your reply.But when I change the num_pts, an error is reported because SMPL initializes with only 6890 points in the canonical space,which means the paras 'xyz' only contains 6890 points.

from gauhuman.

Hi, when you change the number of num_pts, you also need to reset the number of paras 'xyz'. In our case, we randomly set the paras 'xyz' for an ablation study.

from gauhuman.

Hi, when you change the number of num_pts, you also need to reset the number of paras 'xyz'. In our case, we randomly set the paras 'xyz' for an ablation study.

Thanks for your reply. Could you show your code for initialize 'xyz'? Because in your code, you're initializing 'xyz' with 6890 points of SMPL's BigPose.

from gauhuman.

You can use the following line to achieve what you want.
xyz = np.random.random((num_pts, 3)) * (train_cam_infos[0].big_pose_world_bound[1] - train_cam_infos[0].big_pose_world_bound[0]) + train_cam_infos[0].big_pose_world_bound[0]

from gauhuman.