ros 开源物体检测_NO. 1 物体位姿估计 & 机器人抓取
开源算法
【MIT & Princeton 做的物体抓取算法】
基本思想：基于深度学习物体语义分割 -> 基于事先建好的模型进行ICP匹配获得位姿
特色：利用自监督学习减少数据的依赖性
http://apc.cs.princeton.edu/
【TECHNISCHE UNIVERSITAT DRESDEN】
http://cvlab-dresden.de/research/scene-understanding/pose-estimation/#DSAC
【ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes】
http://www.scan-net.org/
【RSS 2018 & code】 PoseCNN: A Convolutional Neural Network for 6D Object Pose Estimation in Cluttered Scenes
https://github.com/yuxng/PoseCNN
【2018MIT的成果，对应伯克利的Dexnet2.0】 Dense Object Nets: Learning Dense Visual Object Descriptors By and For Robotic Manipulat
https://github.com/RobotLocomotion/pytorch-dense-correspondence
【ICRA 2018】Improving 6D Pose Estimation of Objects in Clutter via Physics-aware Monte Carlo Tree Search
概述：基本思路是CNN图像分割+CAD模型匹配，论文主要在CAD模型匹配时提出了Monte Carlo树搜索提升效果
https://github.com/cmitash/PhysimGlobalPose
【CVPR 2019】 PVNet: Pixel-wise Voting Network for 6DoF Pose Estimation
https://github.com/zju3dv/pvnet
【Paper】End-to-End CAD Model Retrieval and 9DoF Alignment in 3D Scans, Technical University of Munich
【 仅需3D模型的高效视觉系统机器手控制】 CRAVES: Controlling Robotic Arm with a Vision-based, Economic System
https://craves.ai/
【CVPR 2019】Segmentation-driven 6D Object Pose Estimation
https://github.com/cvlab-epfl/segmentation-driven-pose
【点云配准网络】 AlignNet-3D for Fast Point Cloud Registration of Partially Observed Objects
Computer Vision
【点云抓握姿态检测】Grasp Pose Detection (GPD)
https://github.com/atenpas/gpd
【目标姿态估计文献资源列表】
https://github.com/YoungXIAO13/ObjectPoseEstimationSummary
【视觉伺服开源平台VISP】支持ROS中使用
http://visp.inria.fr/
【 机器人视觉抓取文献/代码列表】
https://github.com/GeorgeDu/vision-based-robotic-grasping
机器人碰撞检测
通常使用Bounding Box来对机器人抓手、目标物体等进行包围，这样可以有效简化碰撞过程的计算，这类算法通常可以被分为三种基本的bounding box
经典的CODE：
【FCL】
C++: https://github.com/flexible-collision-library/fcl
python: https://github.com/BerkeleyAutomation/python-fcl
中文介绍：ROS(二）FCL碰撞检测算法实现过程_Studying……-CSDN博客_ros 怎么检测机器人存在碰撞
原版论文：http://gamma.cs.unc.edu/FCL/fcl_docs/webpage/pdfs/fcl_icra2012.pdf
【SDK】
C++: https://github.com/bulletphysics/bullet3
python example:
https://github.com/bulletphysics/bullet3/tree/master/examples/pybullet
参考文献：
Daniel Schneider, Elmar Schömer, Nicola Wolpert(2017), “Collision detection for 3D rigid body motion planning with narrow passages“, 2017 IEEE International Conference on Robotics and Automation (ICRA), DOI: 10.1109/ICRA.2017.7989503
Jia Pan, Ioan A. Şucan, Sachin Chitta, Dinesh Manocha(2013), “Real-time collision detection and distance computation on point cloud sensor data”, 2013 IEEE International Conference on Robotics and Automation, DOI: 10.1109/ICRA.2013.6631081
Yue Shen, Qingxuan Jia, Gang Chen, Yifan Wang, Hanxu Sun(2015), “Study of rapid collision detection algorithm for manipulator” 2015 IEEE 10th Conference on Industrial Electronics and Applications (ICIEA) DOI: 10.1109/ICIEA.2015.7334244
技术文章
【直接端到端的深度学习抓取分析】
机器人抓取的前沿到哪了？
【机器人抓取Benchmark】
https://docs.google.com/document/d/1biE0Jmh_5nq-6Giyf2sWZAAQz23uyxhTob2Uz4BjR_w/edit
https://github.com/lianghongzhuo/PointNetGPD
相关赛事
- Amazon Picking Challenge
2. 京东挑战赛
跟踪团队
1.Berkeley的大牛 Sergey Levine团队
http://people.eecs.berkeley.edu/~svlevine/
2. University of Washington and NVIDIA Research
http://rse-lab.cs.washington.edu/