search

Andy97 / DeepMLS

Deep Implicit Moving Least-Squares Functions for 3D Reconstruction
MIT License
130 stars 22 forks source link

what does the ground truth shape mean? #4

Closed THHHomas closed 2 years ago

THHHomas commented 3 years ago

Dear Andy, In your paper, "each input point cloud contains 3000 points, randomly sampled from the ground-truth shape and Gaussian noise with standard deviation set to 0.005 of the maximum bounding box side length of the shape."

I want to ask what you mean "the ground truth shape". The original CAD model in obj file or the mesh from 256**3 voxel followed by the marching cube. Thanks a lot. @Andy97

Andy97 commented 2 years ago

Dear Andy, In your paper, "each input point cloud contains 3000 points, randomly sampled from the ground-truth shape and Gaussian noise with standard deviation set to 0.005 of the maximum bounding box side length of the shape."

I want to ask what you mean "the ground truth shape". The original CAD model in obj file or the mesh from 256**3 voxel followed by the marching cube. Thanks a lot. @Andy97

Hi THHHomas, the "ground truth shape" here is "the mesh from 256**3 voxel followed by the marching cube."

Since some of the original shape models in ShapeNet are troublesome, some shapes are not closed and some polygon faces are flipped, it's not easy to tell whether the given point lies inside or outside the shape, thus we can not compute the signed distance function of shape.

Authors from Occupancy Networks using tsdf fusion to create an fixed version of each shape. We treat this shape as groundtruth shape of each model and compute signed distance function using it and sample points from it to simulate the scanning process in real world.