¡Hola, Mundo!

Problem with earlier methods - PointNet → ignores local geometry- PointNet++ → captures local structure but max pooling → too simple, loses informationThe paper proposesInstead of : neighbor features → max pooling → outputThey do : neighbor features + geometry → attention → weighted sumThe structure of RandLA-Net Local Spatial Encoding (LocSE) ① Input : (N,3+d)- N = number of points- 3 = XYZ coo..

What led to the development of PointNet++ - Pointnet has a disadvantage in that it cannot effectively extract local features because it handles one point independently. (See Full Connected Layer (FC Layer)- PointNet did not reflect the information of the metric space in which the point exists when learning the local structure. Pointnet++ attempted to supplement this point to increase the recogni..

Point Cloud - It is a set of points sprerad in a 3-dimensional space.- A point can be expressed as x, y, and z values.- It has irregular characteristics meaning that the density of points is not uniform.Various ways to express the 3D object. Various ways to express the 3D object. - Point Cloud : A set of points sprerad in a 3-dimensional space.- Voxel(Volumetric Pixel) : 3D model that is expre..

Iterative closest point (ICP) is a point cloud registration algorithm employed to minimize the difference between two clouds of points. Here are two different types of data. 'Model' is the model we are actually looking for, and 'sample' is the result of extracting the data. The purpose of this algorithm is to match 'sample' to 'model'.Somewhere, 'sample' is above a certain coordinate system, a..

It is an iterative method to estimate parameters of a mathematical model from a set of observed data that contains outliers, when outliers do not affect the values of the estimates. Therefore, it also can be interpreted as an outlier detection method. It uses repeated random sub-sampling. If the data itself is completely distorted and not just a little noise as shown in the figure above, so man..