跟踪SLAM前沿动态系列之ICCV2019

跟踪SLAM前沿动态系列之ICCV2019

本文提供的总结和分类，筛选自ICCV2019中与SLAM相关内容，若有遗漏，欢迎评论补充！

注：由于论文开源，泡泡就不提供那个容易失效的网盘分享链接了。另外，具有公开代码的论文，题目已经被标注好颜色咯！

SLAM

1. EMPNet: Neural Localisation and Mapping Using Embedded Memory Points

2. EM-Fusion: Dynamic Object-Level SLAM With Probabilistic Data 

3. Learning Meshes for Dense Visual SLAM

4. ClusterSLAM: A SLAM Backend for Simultaneous Rigid Body Clustering and Motion Estimation    

5. Elaborate Monocular Point and Line SLAM with Robust Initialization    

6. Sequential Adversarial Learning for Self-Supervised Deep Visual Odometry    

7. Distilling Knowledge From a Deep Pose Regressor Network    

8. GSLAM: A General SLAM Framework and  Benchmark (https://github.com/zdzhaoyong/GSLAM）

9. Unsupervised Collaborative Learning of Keyframe Detection and Visual Odometry Towards Monocular Deep SLAM    

自动驾驶

1. Scalable Place Recognition Under Appearance Change for Autonomous Driving    

2. Cross-View Policy Learning for Street   Navigation 

3. Bayesian Relational Memory for Semantic Visual Navigation   

三维重建

1. X-Section: Cross-Section Prediction for Enhanced RGB-D Fusion    

2. Pix2Vox: Context-aware 3D Reconstruction from Single and Multi-view Images

3. Learning to Reconstruct 3D Manhattan Wireframes from a Single Image    

深度估计

1. SynDeMo: Synergistic Deep Feature Alignment for Joint Learning of Depth and Ego-Motion

2. Mono-SF: Multi-View Geometry Meets Single-View Depth for Monocular Scene Flow Estimation of Dynamic Traffic Scenes

3. Digging Into Self-Supervised Monocular Depth Estimation (www.github.com/nianticlabs/monodepth2)

4. Multi-View Stereo by Temporal Nonparametric Fusion (https://aaltoml.github.io/GP-MVS)

5. Learning Joint 2D-3D Representations for Depth Completion

6. Self-Supervised Deep Depth Denoising (https://github.com/VCL3D/DeepDepthDenoising)

7. A Neural Network for Detailed Human Depth Estimation from a Single Image 

8. How Do Neural Networks See Depth in Single Images?

9. Perceptual Deep Depth Super-Resolution

10. 
    

11. Self-Supervised Monocular Depth Hints (www.github.com/nianticlabs/depth-hints)

12. Depth Completion from Sparse LiDAR Data with Depth-Normal Constraints

13. Unsupervised High-Resolution Depth Learning From Videos With Dual Networks

定位

1. CamNet: Coarse-to-Fine Retrieval for Camera Re-Localization

2. TextPlace: Visual Place Recognition and Topological Localization Through Reading Scene Texts 

3. Expert Sample Consensus Applied to Camera Re-Localization    

4. Prior Guided Dropout for Robust Visual  Localization in Dynamic Environments (https://github.com/zju3dv/RVL-Dynamic) 

5. Hierarchical Encoding of Sequential Data with Compact and Sub-linear Storage Cost (https://github.com/intellhave/HESSL)

6. Closed-Form Optimal Two-View Triangulation Based on Angular Errors    

7. Stochastic Attraction-Repulsion Embedding for Large Scale Image Localization (https://github.com/Liumouliu/deepIBL)

8. An Efficient Solution to the Homography-Based Relative Pose Problem With a Common Reference Direction  

9. Privacy Preserving Image Queries for Camera Localization    

10. K-Best Transformation Synchronization    

11. Is This the Right Place? Geometric-Semantic Pose Verification for Indoor Visual Localization (http://www.ok.sc.e.titech.ac.jp/res/RIGHTP/)

12. UprightNet: Geometry-Aware Camera Orientation Estimation from Single Images    

13. Local Supports Global: Deep Camera Relocalization with Sequence Enhancement    

14. SANet: Scene Agnostic Network for Camera   Localization https://github.com/sfu-gruvi-3dv/sanet_relocal_demo

配准    

1. Efficient and Robust Registration on the 3D Special Euclidean Group    

2. Accelerated Gravitational Point Set Alignment with Altered Physical Laws* 

3. Floorplan-Jigsaw: Jointly Estimating Scene Layout and Aligning Partial Scans 

4. DeepVCP: An End-to-End Deep Neural Network for Point Cloud Registration 

5. Efficient and Robust Registration on the 3D Special Euclidean Group    

6. Deep Closest Point: Learning Representations for Point Cloud Registration 

7. Robust Variational Bayesian Point Set Registration 

特征点检测

1. Key.Net: Keypoint Detection by Handcrafted and Learned CNN Filters    
   

2. ELF: Embedded Localisation of Features in Pre-Trained CNN (https://github.com/abenbihi/elf)   

3. Beyond Cartesian Representations for Local Descriptors (https://github.com/cvlab-epfl/log-polar-descriptors)   

4. USIP: Unsupervised Stable Interest Point Detection from 3D Point Clouds (https://github.com/lijx10/USIP) 

5. End-to-End Wireframe Parsing (https://github.com/zhou13/lcnn.) 

目标检测、分割    

1. M3D-RPN: Monocular 3D Region Proposal Network for Object Detection (http://cvlab.cse.msu.edu/project-m3d-rpn.html)

2. Rescan: Inductive Instance Segmentation for Indoor RGBD Scans    

3. 3D Instance Segmentation via Multi-Task Metric Learning    

4. Fine-Grained Segmentation Networks: Self-Supervised Segmentation for Improved Long-Term Visual Localization (https://github.com/maunzzz/fine-grained-segmentation-networks)

5. Explaining the Ambiguity of Object Detection and 6D Pose From Visual Data    

6. Incremental Class Discovery for Semantic Segmentation with RGBD Sensing    

7. Robust Motion Segmentation From Pairwise Matches  

8. Event-Based Motion Segmentation by Motion Compensation    

9. RIO: 3D Object Instance Re-Localization in Changing Indoor Environments Improved Long-Term Visual Localization    

数据集

1. SemanticKITTI: A Dataset for Semantic Scene Understanding of LiDAR Sequences (www.semantic-kitti.org)   

2. Habitat: A Platform for Embodied AI Research    

3. WoodScape: A multi-task, multi-camera fisheye dataset for autonomous driving  (https://github.com/valeoai/WoodScape)    

其他  

1. Neural-Guided RANSAC: Learning Where to Sample Model Hypotheses (vislearn.de/research/neural-guided-ransac/)   

2. Polarimetric Relative Pose Estimation    

3. Quasi-globally Optimal and Efficient Vanishing Point Estimation (https://sites.google.com/view/haoangli/projects/iccv-vp)   

4. Convex Relaxations for Consensus and Non-Minimal Problems in 3D Vision    

5. Homography from two orientation-and scale-covariant features (https://github.com/danini/homography-from-sift-features)

6. Progressive-X: Efficient, Anytime, Multi-Model Fitting Algorithm (https://github.com/danini/progressive-x)

7. Consensus Maximization Tree Search Revisited  (https://github.com/ZhipengCai/MaxConTreeSearch)

8. PLMP - Point-Line Minimal Problems in Complete Multi-View Visibility    

9. Estimating the Fundamental Matrix Without Point Correspondences With Application to Transmission Imaging    

10. A Quaternion-based Certifiably Optimal Solution    

11. Pareto Meets Huber: Efficiently Avoiding Poor Minima in Robust Estimation    

12. Learning Two-View Correspondences and Geometry Using Order-Aware Network    

13. End-to-End Learning of Representations for Asynchronous Event-Based Data (https://github.com/uzh-rpg/rpg_event_representation_learning