Greetings
I've been trying spconv library for point cloud data. I have checked several answers in the issues and the documentation but I am still unsure how the SparseConvTensor is created from the generated voxel.
`
gen = Point2VoxelGPU3d(
vsize_xyz=[0.1, 0.1, 0.1],
coors_range_xyz=[-80, -80, -2, 80, 80, 6],
num_point_features=4, # here num_point_features must equal to pc.shape[1]
max_num_voxels=5000,
max_num_points_per_voxel=5
)
pc = np.random.uniform(-10, 10, size=[100000, 4]).astype(np.float32)
pc_tv = tv.from_numpy(pc).cuda()
voxels_tv, indices_tv, num_p_in_vx_tv = gen.point_to_voxel_hash(pc_tv)
`
I need some clarity on what to do to create sparse tensor after this step. I appreciate your attention and helps. Thank you.
Greetings
I've been trying spconv library for point cloud data. I have checked several answers in the issues and the documentation but I am still unsure how the
SparseConvTensoris created from the generated voxel.`
gen = Point2VoxelGPU3d(
vsize_xyz=[0.1, 0.1, 0.1],
coors_range_xyz=[-80, -80, -2, 80, 80, 6],
num_point_features=4, # here num_point_features must equal to pc.shape[1]
max_num_voxels=5000,
max_num_points_per_voxel=5
)
pc = np.random.uniform(-10, 10, size=[100000, 4]).astype(np.float32)
pc_tv = tv.from_numpy(pc).cuda()
voxels_tv, indices_tv, num_p_in_vx_tv = gen.point_to_voxel_hash(pc_tv)
`
I need some clarity on what to do to create sparse tensor after this step. I appreciate your attention and helps. Thank you.