object-3D目标检测算法调研(基于激光雷达、kitti数据集)
数据来源:kitti数据集及3d-object 竞赛
链接:http://www.cvlibs.net/datasets/kitti/eval_object.php?obj_benchmark=3d
前五名算法简介
截图:
- 1st:
- 名称:VoxelNet++ - LiDAR Only
- 输入数据:LIDAR(only)
- 方法简介:only lidar data(3712 examples) is used. based on my unofficial voxelnet implementation with some improvements
- 基础算法:vocelnet(with some improvement)
- 开源代码:https://github.com/qianguih/voxelnet
- 基础算法简介:
- 网络结构:
- -
- 2ed:
- 名称:MDC
- 方法简介:maskrcnn + ensemble of pointnet
- 输入数据:未说明
- 基础算法:maskrcnn与pointnet结合
- 开源代码:无
- 基础算法简介:暂无
- 3rd:
- 名称:Aggregate View Object Detection with Feature Pyramid Network [la] [AVOD-FPN]
- 方法简介:AVOD+FPN
- 输入数据:LIDAR+RGB
- 开源代码:http://www.cvlibs.net/datasets/kitti/eval_object_detail.php?&result=acde8ce60e154289ada44ecd19697a0327876d1c
- 论文地址:https://arxiv.org/pdf/1711.06396.pdf
- 算法简介:
- 4th:
- 名称:Frustum PointNets for 3D Object Detection from RGB-D Data [la] [F-PointNet]
- 方法简介:
- 输入数据:RGB-D
- 开源模型:https://github.com/charlesq34/frustum-pointnets
- 论文:http://stanford.edu/~rqi/frustum-pointnets/
- 算法简介:-
- 5th:
- 名称:MO3D
- 算法简介:
非深度学习方法:
https://blog.csdn.net/FireMicrocosm/article/details/78059151