【CVPR 2017】PointNet：Deep Learning on Point Sets for 3D Classification and Segmentation

1. Four questions

1. What problem to solve
   Deep learning directly processes 3D point clouds to complete classification and segmentation

2. What method is used to solve
   point-wise + symmetric function (maxpooling)
   

3. How does
   it work
   ? Classification results on ModelNet40 - 89.2
   
   Segmentation results on ShapeNet part dataset. - 83.7
   

4. What is
   the problem? Model each point directly, without more consideration of the connection between points

2. Introduction to the thesis

3. References

Interpretation of PointNet series papers : https://zhuanlan.zhihu.com/p/44809266

PointNet paper notes on 3D classification and segmentation : https://zhuanlan.zhihu.com/p/73086704

PointNet study notes (1) - paper : https://blog.csdn.net/ShuqiaoS/article/details/82983696

Small details: the arxiv version will be more detailed

PointNet paper reproduction and code detailed explanation : https://zhuanlan.zhihu.com/p/86331508

PointNet: A Deep Learning Model for 3D Point Set Classification and Segmentation – with author report

4. Harvest