3D目标检测集合一、基于图像二、基于雷达

78 阅读 0 评论 52 点赞

我是靠谱客的博主正直自行车，最近开发中收集的这篇文章主要介绍3D目标检测集合一、基于图像二、基于雷达，觉得挺不错的，现在分享给大家，希望可以做个参考。

概述

3D目标检测按输入数据类型来分，可分为
单模：激光雷达或相机
多模：激光雷达+相机+其他传感器等

按特征提取来分，可分为：
点云：直接对原始点云信息进行特征提取
体素：将点云通过网格的方式进行划分，随后提取网格特征
鸟瞰图：将3D的信息投影到2D平面中，通常采用鸟瞰图(BEV)的视角，随后使用成熟的2D卷积网络进行特征的提取
图：利用图的方式，对半径R内的点建立图，随后进行特征提取

按训练数据类型来分，可分为：
仅图片
伪点云
点云
xx和xx融合信息
深度值
三维立体信息
关键点

按处理方法来分：
几何先验
额外卷积网络

一、基于图像

1.不需要额外数据输入

RTM3D（2020）
RTM3D: Real-time Monocular 3D Detection from Object Keypoints for Autonomous Driving
论文地址：https://arxiv.org/pdf/2001.03343.pdf
代码地址：https://github.com/Banconxuan/RTM3D

Deep3DBox（2017）
3D Bounding Box Estimation Using Deep Learning and Geometry
论文地址：https://arxiv.org/abs/1612.00496
代码地址：https://github.com/skhadem/3D-BoundingBox

GS3D（2019）
Gs3d: An efficient 3d object detection framework for autonomous driving
论文地址：https://arxiv.org/abs/1903.10955
代码地址：

MonoGRNet（2020）
Monogrnet: A geometric reasoning network for monocular 3d object localization.
论文地址：https://arxiv.org/pdf/1811.10247.pdf
代码地址：https://github.com/Zengyi-Qin/MonoGRNet

FQNet
（相关参考较少）
Deep Fitting Degree Scoring Network for Monocular 3D Object Detection
论文地址：https://arxiv.org/abs/1904.12681
代码地址：

M3D-RPN（2019）
M3D-RPN:Monocular 3D Region Proposal Network for Object Detection
论文地址：https://arxiv.org/abs/1907.06038v1
代码地址：https://github.com/garrickbrazil/M3D-RPN

Shift R-CNN
（相关参考较少）
Shift R-CNN: Deep Monocular 3d Object Detection With Closed-Form Geometric Constraints
论文地址：https://arxiv.org/pdf/1905.09970.pdf
代码地址：

MonoPSR
Monocular 3d object detection leveraging accurate proposals and shape reconstruction

SS3D
Monocular 3d object detection and box fitting trained end-to-end using intersection-over-union loss

M3DSSD
M3DSSD: Monocular 3D Single Stage Object Detector

Movi3D（2020）
Towards generalization across depth for monocular 3d object detection
论文地址：https://arxiv.org/abs/1912.08035v3

OFTNet
Orthographic Feature Transform for Monocular 3D Object Detection
论文地址：https://arxiv.org/abs/1811.08188
代码地址：https://github.com/tom-roddick/oft

ROI-10D
ROI-10D: Monocular Lifting of 2D Detection to 6D Pose and Metric Shape

MonoDIS
Disentangling monocular 3d object detection.

IAFA

MonoDLE

MonoRCNN

Ground-Aware

PCT

MonoGeo

MonoEF

MonoFlex

MonoCon

Task-Aware Monocular Depth Estimation for 3D Object Detection
（分离前景、背景）

Monocular 3D Object Detection and Box Fitting Trained End-to-End Using Intersection-over-Union Loss
（相关参考较少）

Disentangling Monocular 3D Object Detection

Monocular 3D Object Detection via Geometric Reasoning on Keypoints
（相关参考较少）

Monocular 3D Object Detection Leveraging Accurate Proposals and Shape Reconstruction

Accurate Monocular Object Detection via Color-Embedded 3D Reconstruction for Autonomous Driving（将RGB图与点云信息结合）

Towards Generalization Across Depth for Monocular 3D Object Detection

Monocular 3D Detection with Geometric Constraints Embedding and Semi-supervised Training

SMOKE
SMOKE: Single-Stage Monocular 3D Object Detection via Keypoint Estimation
论文地址：https://arxiv.org/pdf/2002.10111.pdf
代码地址：https://github.com/lzccccc/SMOKE

FCOS3D
FCOS3D: Fully Convolutional One-Stage Monocular 3D Object Detection
论文地址：https://arxiv.org/abs/2104.10956
代码地址：https://github.com/open-mmlab/mmdetection3d

DD3D
Is Pseudo-Lidar needed for Monocular 3D Object detection?
论文地址：https://arxiv.org/abs/2108.06417
代码地址：https://github.com/TRI-ML/dd3d

GUP Net
Geometry Uncertainty Projection Network for Monocular 3D Object Detection
论文地址：https://arxiv.org/abs/2107.13774
代码地址：https://github.com/SuperMHP/GUPNet

MonoPair
MonoPair: Monocular 3D Object Detection Using Pairwise Spatial Relationships
论文地址：https://arxiv.org/abs/2003.00504
代码地址：

PGD
论文地址：https://arxiv.org/abs/2107.14160
代码地址：https://github.com/open-mmlab/mmdetection3d

DETR3D
DETR3D: 3D Object Detection from Multi-view Images via 3D-to-2D Queries
论文地址：https://arxiv.org/abs/2110.06922
代码地址：https://github.com/WangYueFt/detr3d

Monocular 3D Object Detection with Decoupled Structured Polygon Estimation and Height-Guided Depth Estimation

Object-Aware Centroid Voting for Monocular 3D Object Detection

Learning Monocular 3D Vehicle Detection without 3D Bounding Box Labels

Depth-conditioned Dynamic Message Propagation for Monocular 3D Object Detection

3DIoUMatch: Leveraging IoU Prediction for Semi-Supervised 3D Object Detection

Categorical Depth Distribution Network for Monocular 3D Object Detection

Geometry-based Distance Decomposition for Monocular 3D Object Detection

Geometry-aware data augmentation for monocular 3D object detection

Multi-View Reprojection Architecture for Orientation Estimation.

Monocular 3D Object Detection Leveraging Accurate Proposals and Shape Reconstruction.

ROI-10D: Monocular Lifting of 2D Detection to 6D Pose and Metric Shape

DSGN
（双目）
DSGN: Deep Stereo Geometry Network for 3D Object Detection
论文地址：https://arxiv.org/pdf/2001.03398.pdf
代码地址：https://github.com/Jia-Research-Lab/DSGN

Learning Depth-Guided Convolutions for Monocular 3D Object Detection

Kinematic 3D Object Detection in Monocular Video

Rethinking Pseudo-LiDAR Representation.

RTS3D: Real-time Stereo 3D Detection from 4D Feature-Consistency Embedding Space for Autonomous Driving

Objects are Different: Flexible Monocular 3D Object Detection

GrooMeD-NMS: Grouped Mathematically Differentiable NMS for Monocular 3D Object Detection.

Delving into Localization Errors for Monocular 3D Object Detection

Depth-conditioned Dynamic Message Propagation for Monocular 3D Object Detection

M3DSSD: Monocular 3D Single Stage Object Detector

MonoRUn: Monocular 3D Object Detection by Self-Supervised Reconstruction and Uncertainty Propagation

Categorical Depth Distribution Network for Monocular 3D Object Detection

Monocular 3D Object Detection: An Extrinsic Parameter Free Approach

AutoShape: Real-Time Shape-Aware Monocular 3D Object Detection

Geometry-based Distance Decomposition for Monocular 3D Object Detection

Are we Missing Confidence in Pseudo-LiDAR Methods for Monocular 3D Object Detection?

Ground-Aware Monocular 3D Object Detection for Autonomous Driving

Learning Auxiliary Monocular Contexts Helps Monocular 3D Object Detection

Diversity Matters: Fully Exploiting Depth Clues for Reliable Monocular 3D Object Detection

Pseudo-Stereo for Monocular 3D Object Detection in Autonomous Driving

MonoJSG: Joint Semantic and Geometric Cost Volume for Monocular 3D Object Detection

Time3D: End-to-End Joint Monocular 3D Object Detection and Tracking for Autonomous Driving

Dimension Embeddings for Monocular 3D Object Detection

Homography Loss for Monocular 3D Object Detection

Exploring Geometric Consistency for Monocular 3D Object Detection

MonoDTR: Monocular 3D Object Detection with Depth-Aware Transformer

PersDet: Monocular 3D Detection in Perspective Bird’s-Eye-View

Densely Constrained Depth Estimator for Monocular 3D Object Detection

DID-M3D: Decoupling Instance Depth for Monocular 3D Object Detection

DEVIANT: Depth EquiVarIAnt NeTwork for Monocular 3D Object Detection

Monocular 3D Object Detection with Depth from Motion

2.基于立体信息

Mono3D

3.基于深度信息

3DOP

stereoRCNN

4.基于形状

MF3D

Peseudo-LiDAR

MonoPSR

AM3D

5.基于分割

Mono3D++

Deep-MENTA

3DVP

6.其他

Mono3D
3D Object Detection for Autonomous Driving: A Review and New Outlooks
论文地址：https://arxiv.org/abs/2206.09474
项目地址：https://xiaozhichen.github.io/

SubCNN
SubCNN: Subcategory-aware Convolutional Neural Networks for Object Proposals and Detection
论文地址：https://arxiv.org/abs/1604.04693v3
项目地址：https://github.com/tanshen/SubCNN

二、基于雷达

相关链接参考：
https://paperswithcode.com/sota/monocular-3d-object-detection-on-kitti-cars（kitti数据集3D检测排行榜）
https://www.nuscenes.org/object-detection?externalData=all&mapData=all&modalities=Camera（nuscenes检测任务排行榜）
https://zhuanlan.zhihu.com/p/450995240
https://zhuanlan.zhihu.com/p/57029694