/PointCloud_Classification_using_optimized_PointNet_by_kdtree

Optimized the CNN framework PointNet by using kdtree

Primary LanguagePython

PointCloud_Classification_using_optimized_PointNet

  1. Use SGD not mini-batch like PointNet

  2. By using kdtree, I've solved the two weak points of the PointNet, which is a great open-source framework for point cloud classification： 1> All the point clouds used in PointNet must have the same point number. 2> All the point clouds used in PointNet must be uniformly sampled. These two requirements can not be satisfied in actual situation, which limit the application of the PointNet