1. Four questions

1. What problem to solve

At present self-reconstruction The main learning global geometry, Failed to explore effectively local geometry. Explore unsupervised Study local geometry And the lack of Effective and semantic (semantic) Local structure of (local structure) supervise The signal

Because of lacking effective and semantic local structure supervision, however, error may accumulate in the local structure learning process, which limits the network’s ability in 3D point cloud understanding.

2. What method has been used to solve

Three branches ： stay 3. Overview In detail

• （ Universal ）branch A： use RNN To aggregate all the features
• （ above ）branch R： Rebuild the whole ,global , belt VAE？ Can produce more feature space

Self-reconstruction is started from a variational feature space, which enables MAP-VAE to generate new shapes by capturing the distribution information over training point clouds in the feature space

• （ below ）branch P： Divide the original point cloud input into the front half and the back half ,half-to-half prediction

We introduce multi-angle analysis for point clouds to mine effective local self-supervision, and combine it with global self-supervision under a variational constraint.

3. What's the effect

ModelNet Effect on ：

segmentation on Shape part dataset（ Main contrast LGAN？ I haven't seen it …）

4. What's the problem

？

2. Paper introduction

3. Reference material

paper

4. Harvest

Research motivation ？： Intuitively speaking , Look at the point cloud from different angles , It can clearly show the corresponding relationship and relationship between different shape areas , That is, the correspondence between the front half and the rear half of the shape in each view

Innovation points ：？

• multi-angle： The division of angles also pays attention to , I don't think much about it
• variational constraint：VAE？ Variational encoder ？ Can produce more shape （VAE The role of ）

Provides a new agent task ：1. Divide it in half and predict it in half ;2. Rebuild the whole （ common ）

key word ：semantic, local structure