[轻松入门]3D点云检测/3D雷达检测学习资料(一)

最近自动驾驶是一个小风口，小菜鸡凑巧被mentor分配了研究3d点云检测，于是搭上了这路末班车，但在研究这一块的时候发现：

[图片上传失败...(image-3a5c8a-1649212867710)]

• 3d点云检测目前还处于2d检测领域初中期的阶段，学习资料有点少，导致我初期学的时候不清楚常见的trick有哪些，改哪些模块可能会有效。说的更简单一点，3d检测入门前期快速培养intuition太难，不如2d检测容易。
• 大家更多的关注的都是精度，可能偶尔会有一些零零散散的文章关注速度。初期找这一部分文章花了很久很久。
• 没有一个随时更新的paper list供个人来follow学术前沿。比如现在cvpr2022结果出来了，但还没有人汇总cvpr20223d检测的文章。

鉴于以上原因，小菜鸡在GitHub总结了一个相关资源的list，【如果你觉得有用，请给我点个star，感谢～】

GitHub - TianhaoFu/Awesome-3D-Object-Detection: Papers, code and datasets about deep learning for 3D Object Detection.github.com/TianhaoFu/Awesome-3D-Object-Detection/[图片上传失败...(image-933e40-1649212867706)]

并且会不断维护更新，无他，只是不想自己走过的弯路让别人再走一遍，希望各路大神能前来指点一二。

如何用好这个repo?

1. 大致扫完一眼后，快速进入blog部分，学习经典方法。
2. 进入course部分，学习图宾根大学课程的3d检测部分。
3. 进入video部分，看3d检测相关的seminar。
4. 现在，你已经大致具备3d检测领域的intuition了，之后

• 如果你想发论文，你可以进入paper部分。
• 如果你想打比赛，你可以进入competition solution部分。【等待更新】
• 如果你想做工程，你可以进入engineering部分。【等待更新】

repo内容有哪些？【部分内容节选】

Dataset

• KITTI Dataset

• 3,712 training samples

• 3,769 validation samples

• 7,518 testing samples

• nuScenes Dataset

• 28k training samples

• 6k validation samples

• 6k testing samples

Top conference & workshop

Conferene

• Conference on Computer Vision and Pattern Recognition(CVPR)
• International Conference on Computer Vision(ICCV)
• European Conference on Computer Vision(ECCV)

Workshop

• CVPR 2021 Workshop on Autonomous Driving(waymo 3D detection)
• ICCV 2021 Workshop on Autonomous Vehicle Vision (AVVision), note
• ICCV 2021 Workshop SSLAD Track 2 - 3D Object Detection
• ECCV 2020 Workshop on Perception for Autonomous Driving

Paper (Lidar-based method)

• HVNet: Hybrid Voxel Network for LiDAR Based 3D Object Detection（CVPR2020) paper
• LiDAR R-CNN: An Efficient and Universal 3D Object Detector（CVPR2021) paper
• Center-based 3D Object Detection and Tracking(CVPR2021) paper
• 3DIoUMatch: Leveraging IoU Prediction for Semi-Supervised 3D Object Detection(CVPR2021) paper
• Embracing Single Stride 3D Object Detector with Sparse Transformer(CVPR2022) paper, code
• Point Density-Aware Voxels for LiDAR 3D Object Detection(CVPR2022) paper, code
• A Unified Query-based Paradigm for Point Cloud Understanding(CVPR2022) paper
• Beyond 3D Siamese Tracking: A Motion-Centric Paradigm for 3D Single Object Tracking in Point Clouds(CVPR2022) paper, code
• Not All Points Are Equal: Learning Highly Efficient Point-based Detectors for 3D LiDAR Point Clouds(CVPR2022) paper, code
• Back To Reality: Weakly-supervised 3D Object Detection with Shape-guided Label Enhancement(CVPR2022) paper, code
• Voxel Set Transformer: A Set-to-Set Approach to 3D Object Detection from Point Clouds(CVPR2022) paper, code

Survey

• 2021.04 Point-cloud based 3D object detection and classification methods for self-driving applications: A survey and taxonomy paper
• 2021.07 3D Object Detection for Autonomous Driving: A Survey paper
• 2021.07 Multi-Modal 3D Object Detection in Autonomous Driving: a Survey paper
• 2021.10 A comprehensive survey of LIDAR-based 3D object detection methods with deep learning for autonomous driving paper
• 2021.12 Deep Learning for 3D Point Clouds: A Survey paper

Book

• 3D Object Detection Algorithms Based on Lidar and Camera: Design and Simulation book

Video

• Aivia online workshop: 3D object detection and tracking video
• 3D Object Retrieval 2021 workshop video
• 3D Deep Learning Tutorial from SU lab at UCSD video
• Lecture: Self-Driving Cars (Prof. Andreas Geiger, University of Tübingen) video
• Current Approaches and Future Directions for Point Cloud Object (2021.04) video
• Latest 3D OBJECT DETECTION with 30+ FPS on CPU - MediaPipe and OpenCV Python (2021.05) video

Course

• University of Toronto, csc2541
• University of Tübingen, Self-Driving Cars(Strong Recommendation)

Blog

• Waymo Blog
• PointNet系列论文解读
• Deep3dBox: 3D Bounding Box Estimation Using Deep Learning and Geometry
• SECOND算法解析
• PointRCNN深度解读
• Fast PointRCNN论文解读
• PointPillars论文和代码解析
• VoxelNet论文和代码解析
• CenterPoint源码分析
• PV-RCNN: 3D目标检测 Waymo挑战赛+KITTI榜 单模态第一算法
• LiDAR R-CNN：一种快速、通用的二阶段3D检测器
• 混合体素网络（HVNet)
• 自动驾驶感知| Range Image paper分享
• SST：单步长稀疏Transformer 3D物体检测器

Famous Research Group/Scholar

• Naiyan Wang@Tusimple
• Hongsheng Li@CUHK
• Oncel Tuzel@Apple
• Oscar Beijbom@nuTonomy
• Raquel Urtasun@University of Toronto
• Philipp Krähenbühl@UT Austin
• Deva Ramanan@CMU
• Jiaya Jia@CUHK
• Thomas Funkhouser@princeton
• Leonidas Guibas@Stanford
• Steven Waslander@University of Toronto
• Ouais Alsharif@Google Brain
• Yuning CHAI(former)@waymo

Famous CodeBase

• Point Cloud Library (PCL)
• Spconv
• Det3D
• mmdetection3d
• OpenPCDet
• Centerpoint

Famous Toolkit

• ZED Box

以上就是一个小小的简短的介绍，未完待续，之后的文章将具体讲一些3d检测领域的paper套路、比赛套路、工程套路等等。【自己总结的肯定和各路大神比不了，所以非常非常欢迎各路大神能前来和小菜鸡交流】
希望大家多多多提一些意见，共同进步！