1. SECOND & History

paper:SECOND: Sparsely Embedded Convolutional Detection

graph TB
    VoxelNet --> SECOND --> PointPillar

Related paper: VoxelNet: stacked VFE(voxel feature encoding module) + 3D Conv RPN This paper:

3D Conv --> Sparse Conv --> improved sparse conv
angle loss regression approach
  1. We apply sparse convolution in LiDAR-based object detection, thereby greatly increasing the speeds of training and inference.
  2. We propose an improved method of sparse convolution that allows it to run faster.
  3. We propose a novel angle loss regression approach that demonstrates better orientation regression performance than other methods do.
  4. We introduce a novel data augmentation method for LiDAR-only learning problems that greatly increases the convergence speed and performance.

2. Pipeline

graph TB
    VoxelNet_BackBone[Voxelization, Sampling and stacked VFE like VoxelNet] --> encoded_features(Encoded Features) --> SparseConv --> RPN

3. Details

3.1 Improved Sparse Convolution

Data Augmentation

Ground Truth database is generated that contains the attributes of objs and associated pts cloud data.