3D Convolutional Neural Networks

This is an extension on Convolutional Neural Network, focusing on 3D convolutional neural network applications and technologies.

01 Background

Datasets

KITTI Dataset ModelNet40

• https://modelnet.cs.princeton.edu ScanNet
• http://www.scan-net.org TU-berlin Quick Draw
• https://github.com/googlecreativelab/quickdraw-dataset MNIST

CIFAR10

Resources

https://medium.com/mlearning-ai/an-introduction-to-3d-object-detection-in-autonomous-navigation-3a8779de6443

https://paperswithcode.com/sota/3d-point-cloud-classification-on-modelnet40

02 Important Concepts

3D Representations

Approaches for 3D Object Detection

• Point-wise feature extraction
• Aggregate view object detection
• Frustum / Region of Interest

03 Famous Architectures

PointNet (2017)

📃: 💡: Resources:

• https://medium.com/@luis_gonzales/an-in-depth-look-at-pointnet-111d7efdaa1a

VoxelNet (2017)

📃: 💡:

PointCNN (2018)

📃: 💡:

PointRCNN (2019)

📃: PointRCNN: 3D Object Proposal Generation and Detection from Point Cloud 💡: First two-stage 3D object detector for 3D object detection by only using point clouds.