Table 1 Acronyms and notations
FPS | Farthest Point Sampling |
|---|---|
3DEPS | 3D Edge-preserving Sampling |
RS | Random Sampling |
UVS | Uniformly Voxelized Sampling |
VFPS | Voxelized Farthest Point Sampling |
VBS | Voxel-based Sampling |
SBF | 3D Surface Boundary Filter |
LFEOs | Local Feature Extraction Operations |
FFM | Feature Fusion Module |
LFEM | Local Feature Extraction Module |
DGFFM | Dual-granularity Feature Fusion Module |
GT | Ground truth |
TP | True Positive |
FP | False Positive |
FN | False Negative |
IoU | Intersection over Union |
AveDiff | The average difference to the best performer |
mCov | Mean coverage |
mWCov | Mean weighted coverage |
\({\mathcal{P}}\) | The original point set or point cloud |
\(M\) | The number of points in the original \({\mathcal{P}}\) |
\(n\) | The number of points after down-sampling |
\(m\) | The number of remaining points to be sampled |
\(N\) | Number of points not yet visited in \({\mathcal{P}}\) |
\(S(m,N)\) | The random variable used in Random Sampling |
\(F( \cdot )\) | A probability distribution function |
\(A:\) | Permutation |
\(U,V\) | Random variables with a uniform distribution in (0, 1) |
\(A \leftarrow B\) | Assign the value B to A |
\({\mathcal{P}}_{i}\) | The points contained in the i-th voxel |
\(c_{i}\) | The gravity centroid in XYZ space of \({\mathcal{P}}_{i}\) |
\(l_{x}\), \(l_{y}\),\(l_{z}\) | The length, width, and height of each voxel |
\({\mathcal{C}}\) | Internal point set |
\({\mathcal{B}}\) | Edge point set |
\(C\) | The number of semantic classes |
\(C_{ins}\) | The number of semantic classes that have instances |
IoU \(( \cdot, \cdot)\) | Intersection over Union calculation of two entities |