ICRA 2019 论文速览 | 基于Deep Learning 的SLAM

点击上方“计算机视觉life”，选择“星标”

快速获得最新干货

笔者汇总了ICRA 2019 SLAM相关论文，总共分为四个部分：

Deep learning + traditional SLAM，见 ICRA 2019 论文速览 | SLAM 爱上 Deep Learning

Traditional SLAM/3D vision，见ICRA 2019 论文速览 | 传统SLAM、三维视觉算法进展

Deep learning based SLAM

SLAM evaluation and datasets

本文介绍:deep learning based SLAM

后续文章敬请期待

1.E2E-VO/ SLAM

GEN-SLAM: Generative Modeling for Monocular Simultaneous Localization and Mapping（深度学习位姿和深度图）

Keywords: SLAM, Localization, Visual-Based Navigation

Beyond Photometric Loss for Self-Supervised Ego-Motion Estimation（深度学习，自监督的深度和里程计，参考了GeoNet和SfmLearner）

Keywords: SLAM, Visual Learning, Localization

代码：
https://github.com/hlzz/DeepMatchVO

Learning Monocular Visual Odometry through Geometry-Aware Curriculum Learning（深度学习的VO）

Keywords: Localization, Visual Learning, Deep Learning in Robotics and Automation

GANVO: Unsupervised Deep Monocular Visual Odometry and Depth Estimation with Generative Adversarial Networks（深度学习 基于GAN的无监督深度和VO方法）

Keywords: Deep Learning in Robotics and Automation, Localization, Visual Tracking

Unsupervised Learning of Monocular Depth and Ego-Motion Using Multiple Masks（无监督深度学习的深度图和位姿网络）

Keywords: Deep Learning in Robotics and Automation, SLAM

2. E2E Navigation

（AWARD）Variational End-To-End Navigation and Localization（端到端定位导航）

Keywords: Deep Learning in Robotics and Automation, Computer Vision for Transportation, Autonomous Vehicle Navigation

Deep Reinforcement Learning of Navigation in a Complex and Crowded Environment with a Limited Field of View（强化学习机器人视觉导航）

Keywords: Deep Learning in Robotics and Automation, Collision Avoidance, Service Robots

Generalization through Simulation: Integrating Simulated and Real Data into Deep Reinforcement Learning for Vision-Based Autonomous Flight（强化学习的无人机自主导航）

Keywords: Deep Learning in Robotics and Automation

代码
https://github.com/gkahn13/GtS

3.Feature & VPR

（AWARD）Learning Scene Geometry for Visual Localization in Challenging Conditions （RGB和Depth中找出场景的结构化描述特征用于VPR）

Keywords: Localization, RGB-D Perception, Computer Vision for Other Robotic Applications

Localizing Discriminative Visual Landmarks for Place Recognition（VPR路标的显著性检测）

Keywords: Localization, Visual-Based Navigation, Computer Vision for Automation

Improving Keypoint Matching Using a Landmark-Based Image Representation（深度学习地标区域描述符和特征点匹配）

Keywords: SLAM, Localization

A Comparison of CNN-Based and Hand-Crafted Keypoint Descriptors（传统和深度学习特征描述子的光照和角度变化下的性能分析）

Keywords: SLAM, Visual-Based Navigation, Deep Learning in Robotics and Automation

A Multi-Domain Feature Learning Method for Visual Place Recognition（迁移学习的特征学习用于场景识别）

Keywords: Localization, SLAM, Performance Evaluation and Benchmarking

Night-To-Day Image Translation for Retrieval-Based Localization（图像迁移方法的的位置定位）

Keywords: Localization, Visual Learning, Autonomous Vehicle Navigation

2D3D-MatchNet: Learning to Match Keypoints across 2D Image and 3D Point Cloud（深度学习，2D3D数据下的匹配特征点提取网络）

Feng, Mengdan    National University of Singapore

Keywords: Deep Learning in Robotics and Automation, Visual Learning, Localization

Look No Deeper: Recognizing Places from Opposing Viewpoints under Varying Scene Appearance Using Single-View Depth Estimation（用深度学习的深度预测来完成反向视角下的VPR）

Keywords: Localization, Deep Learning in Robotics and Automation

Multi-Process Fusion: Visual Place Recognition Using Multiple Image Processing Methods——IRAL（图像上多信息融合做VPR）

Keywords: Localization, Visual-Based Navigation

4.Depth & Disparity

（AWARD）Geo-Supervised Visual Depth Prediction（深度图网络）

Keywords: Visual Learning, Sensor Fusion

代码
https://github.com/feixh/GeoSup

FastDepth: Fast Monocular Depth Estimation on Embedded Systems（178fps TX2上的224x224深度图计算方法）

Keywords: Deep Learning in Robotics and Automation, Range Sensing, Computer Vision for Other Robotic Applications

代码
http://fastdepth.mit.edu
https://github.com/dwofk/fast-depth

SuperDepth: Self-Supervised, Super-Resolved Monocular Depth Estimation

Keywords: Deep Learning in Robotics and Automation, Visual Learning, Mapping

Depth Completion with Deep Geometry and Context Guidance（稀疏深度图补齐网络）

Keywords: RGB-D Perception, Computer Vision for Other Robotic Applications

Self-Supervised Sparse-To-Dense: Self-Supervised Depth Completion from LiDAR and Monocular Camera（自监督学习的Lidar深度数据补齐）

Keywords: Visual Learning, RGB-D Perception, Sensor Fusion

代码
https://github.com/fangchangma/self-supervised-depth-completion

Self-Supervised Learning for Single View Depth and Surface Normal Estimation（自监督的深度和法向图估计）

Keywords: Deep Learning in Robotics and Automation, Visual Learning, Mapping

Plug-And-Play: Improve Depth Prediction Via Sparse Data Propagation（循环优化深度图）

Keywords: Deep Learning in Robotics and Automation, RGB-D Perception, Computer Vision for Automation

Depth Generation Network: Estimating Real World Depth from Stereo and Depth Images（左右图生成深度图网络）

Keywords: AI-Based Methods, RGB-D Perception, Range Sensing

Anytime Stereo Image Depth Estimation on Mobile Devices（双目深度图匹配计算方法，快速）

Keywords: Deep Learning in Robotics and Automation, Computer Vision for Automation, Computer Vision for Other Robotic Applications

代码
https://github.com/mileyan/AnyNet

UWStereoNet: Unsupervised Learning for Depth Estimation and Color Correction of Underwater Stereo Imagery（深度学习的双目立体匹配）

Keywords: Marine Robotics, Deep Learning in Robotics and Automation, Computer Vision for Other Robotic Applications

Real-Time Joint Semantic Segmentation and Depth Estimation Using Asymmetric Annotations（深度学习语义和深度的分割网络）

Keywords: Visual Learning, Semantic Scene Understanding, SLAM

代码
https://github.com/DrSleep/multi-task-refinenet

DSNet: Joint Learning for Scene Segmentation and Disparity Estimation（深度学习左右图估计语义分割和深度图）

Keywords: Semantic Scene Understanding, Deep Learning in Robotics and Automation, Object Detection, Segmentation and Categorization

SweepNet: Wide-Baseline Omnidirectional Depth Estimation（宽基线，多摄像头的深度估计方法）

Keywords: Omnidirectional Vision, Computer Vision for Automation, Deep Learning in Robotics and Automation

A Supervised Approach to Predicting Noise in Depth Images（预测深度图的噪声区域）

Keywords: RGB-D Perception, Perception for Grasping and Manipulation, Deep Learning in Robotics and Automation

5. Point Cloud Segmentation

SqueezeSegV2: Improved Model Structure and Unsupervised Domain Adaptation for Road-Object Segmentation from a LiDAR Point Cloud （深度学习，雷达数据分割路面上的物体）

Keywords: Object Detection, Segmentation and Categorization, Semantic Scene Understanding, AI-Based Methods

https://github.com/BichenWuUCB/SqueezeSeg
https://github.com/xuanyuzhou98/SqueezeSegV2

Hierarchical Depthwise Graph Convolutional Neural Network for 3D Semantic Segmentation of Point Clouds（点云语义分割方法）

Keywords: Semantic Scene Understanding, AI-Based Methods, RGB-D Perception

6. Autonomous Vehicle

Learning to Drive from Simulation without Real World Labels（自动驾驶中的学习方法）

Keywords: Deep Learning in Robotics and Automation, Visual Learning, Learning from Demonstration

Learning to Drive in a Day（强化学习的自动驾驶）

Keywords: AI-Based Methods, Deep Learning in Robotics and Automation, Computer Vision for Transportation

Building a Winning Self-Driving Car in Six Months（与Uber 合作的自动驾驶平台）

Keywords: Autonomous Vehicle Navigation, Intelligent Transportation Systems, Computer Vision for Transportation

Multimodal Spatio-Temporal Information in End-To-End Networks for Automotive Steering Prediction （BMW合作的自动驾驶）

Keywords: Autonomous Vehicle Navigation, Deep Learning in Robotics and Automation, Visual Learning

Monocular Semantic Occupancy Grid Mapping with Convolutional Variational Encoder-Decoder Networks——IRAL（单目图生成避障图）

Keywords: Semantic Scene Understanding, Object Detection, Segmentation and Categorization, Computer Vision for Transportation

交流群

欢迎加入公众号读者群一起和同行交流，目前有SLAM、算法竞赛、图像检测分割、人脸人体、医学影像、自动驾驶、综合等微信群（以后会逐渐细分），请扫描下面微信号加群，备注：”昵称+学校/公司+研究方向“，例如：”张三 + 上海交大 + 视觉SLAM“。请按照格式备注，否则不予通过。添加成功后会根据研究方向邀请进入相关微信群。请勿在群内发送广告，否则会请出群，谢谢理解~

从零开始学习三维视觉核心技术SLAM，扫描查看介绍，3天内无条件退款

学习切忌单打独斗，一个良好的学习圈子能够帮助你快速入门，交流讨论才能少走弯路，快速进步！

如有AI领域实习、求职、招聘、项目合作、咨询服务等需求，快来加入我们吧，期待和你建立连接，找人找技术不再难！

推荐阅读

ICRA 2019 论文速览 | SLAM 爱上 Deep Learning

ICRA 2019 论文速览 | 传统SLAM、三维视觉算法进展

吐血整理 | SLAM方向国内有哪些优秀的公司？

最强战队 | 三维视觉、SLAM方向全球顶尖实验室汇总

从零开始一起学习SLAM | 为什么要学SLAM？

从零开始一起学习SLAM | 学习SLAM到底需要学什么？

从零开始一起学习SLAM | SLAM有什么用？

从零开始一起学习SLAM | C++新特性要不要学？

从零开始一起学习SLAM | 为什么要用齐次坐标？

从零开始一起学习SLAM | 三维空间刚体的旋转

从零开始一起学习SLAM | 为啥需要李群与李代数？

从零开始一起学习SLAM | 相机成像模型

从零开始一起学习SLAM | 不推公式，如何真正理解对极约束?

从零开始一起学习SLAM | 神奇的单应矩阵

从零开始一起学习SLAM | 你好，点云

从零开始一起学习SLAM | 给点云加个滤网

从零开始一起学习SLAM | 点云平滑法线估计

从零开始一起学习SLAM | 点云到网格的进化

从零开始一起学习SLAM | 理解图优化，一步步带你看懂g2o代码

从零开始一起学习SLAM | 掌握g2o顶点编程套路

从零开始一起学习SLAM | 掌握g2o边的代码套路

从零开始一起学习SLAM | ICP原理及应用

 最新AI干货，我在看