2021년 5월 SLAM 뉴스

Posted on Edited on In , , Views:

논문 이름 누르면 자세한 정보가 열립니다!

학회

 

이번 달 내가 관심가지는 논문들 (키노트/랜드마크 급)

Monocular Visual Odometry Using Template Matching and IMU - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9427190) - Sebastian Scherer 교수님 랩실에서 나온 논문 - 지상을 바라보는 카메라로 VO를 할 때 template matching을 통해 Ackerman steering model에 대한 pose 변환을 잴 수 있으나, 강건하지 못하다. - IMU를 사용해서 yaw 값을 치환해서 성능을 높인다.
Super Odometry: IMU-centric LiDAR-Visual-Inertial Estimator for Challenging Environments - [논문 링크](https://arxiv.org/pdf/2104.14938.pdf) - Sebastian Scherer 교수님 랩실에서 나온 논문 - IMU를 중심으로 Camera + LiDAR를 퓨전해서 퓨어 비전 / 퓨어 라이다로는 어려운 공간에서 안정적으로 맵핑할 수 있는 기술
Simultaneous Localization and Mapping: A Rapprochement of Filtering and Optimization-Based Approaches - [논문 링크](https://www2.eecs.berkeley.edu/Pubs/TechRpts/2021/EECS-2021-76.pdf) - An invitiation to 3D vision의 저자인 Yi Ma와 Sastry가 마킹한 버클리 졸업논문. - Optimisation 방식과 EKF 방식을 비교 및 짬뽕
Learning to Predict Repeatability of Interest Points - [논문 링크](https://arxiv.org/pdf/2105.03578.pdf) - Niantic 팀과 Tat-jun Chin이 참여 - 시간에 따른 interest point의 repeatability를 평가하는 시스템을 만듬. - 하루의 시간동안 바뀌는 조명에서... - 계절이 바뀌면서...
TSDF++: A Multi-Object Formulation for Dynamic Object Tracking and Reconstruction - [논문 링크](https://arxiv.org/pdf/2105.07468.pdf) - Roland Siegwart, Juan Nieto 교수님 참여 - 하나의 reconstruction volume으로 scene 전체와 모든 object를 표현하는 multi-object TSDF 기법을 소개함. - 기존의 TSDF 방식으로 다수의 object를 표현하기 위해서는 object의 수 만큼 reconstruction volume이 필요했어서 scalable하지 않았음.
Enhancing Photorealism Enhancement - [논문 링크](https://arxiv.org/abs/2105.04619) - Vladlen Koltun + 인텔 자율주행 랩 - 가상 이미지를 실제처럼 보이게 바꿔주는 네트워크를 만듬.
USACv20: robust essential, fundamental and homography matrix estimation - [논문 링크](https://arxiv.org/pdf/2104.05044.pdf) - SAC계열의 고수들: Daniel Barath, Jiri Matas - RANSAC 계열 기술들을 리뷰하고, 가장 좋은것들만 골라서 USAC을 만듬. - (근데 USAC 논문은 바뀐건 크게 없는거 같은데 왜 매번 나오는거지...?)
A Front-End for Dense Monocular SLAM using a Learned Outlier Mask Prior - [논문 링크](https://arxiv.org/pdf/2104.00562.pdf) - John Leonard 교수님 랩실 - Depth from video를 수행하면서 부산물로 생기는 outlier mask까지 SLAM에 사용해서 depth 정보를 더 좋게 만들고, CNN-SLAM 파이프라인에 넣어서 Dense monocular SLAM의 frontend를 구현
The Temporal Opportunist: Self-Supervised Multi-Frame Monocular Depth - [논문 링크](https://arxiv.org/pdf/2104.14540.pdf) - Victor Prisacariu & Niantic 팀 - 학습된 정보 뿐이 아닌 실시간으로 들어오는 정보까지 사용해서 더 좋은 성능을 내는 dense depth estimation 기법을 소개함. - 해당 기법은 self-supervision으로만 학습됨.
드론 자율비행 기술 동향 - [논문 링크](https://www.koreascience.or.kr/article/JAKO202153150644135.pdf) - ETRI
Guest Editorial: Special Issue on Performance Evaluation in Computer Vision - [논문 링크](https://link.springer.com/article/10.1007/s11263-021-01455-x) - Torsten Sattler

 

그 외

MonoGRNet: A General Framework for Monocular 3D Object Detection - [논문 링크](https://arxiv.org/pdf/2104.08797.pdf)
FCOS3D: Fully Convolutional One-Stage Monocular 3D Object Detection - [논문 링크](https://arxiv.org/pdf/2104.10956.pdf)
PLC-VIO: Visual-Inertial Odometry Based on Point-Line Constraints - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9431679)
Scene Coordinate Regression Network With Global Context-Guided Spatial Feature Transformation for Visual Relocalization - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9437699)
Adaptive depth estimation for pyramid multi-view stereo - [논문 링크](http://graphvision.whu.edu.cn/papers/liaojie2021cag.pdf)
Self-Point-Flow: Self-Supervised Scene Flow Estimation from Point Clouds with Optimal Transport and Random Walk - [논문 링크](https://arxiv.org/pdf/2105.08248.pdf)
VS-Net: Voting with Segmentation for Visual Localization - [논문 링크](https://arxiv.org/pdf/2105.10886.pdf)
Improved Real-Time Monocular SLAM Using Semantic Segmentation on Selective Frames - [논문 링크](https://arxiv.org/pdf/2105.00114.pdf)
Visual SLAM based on instance segmentation in dynamic scenes - [논문 링크](https://iopscience.iop.org/article/10.1088/1361-6501/abfceb/meta)
Factor graph optimization for GNSS/INS integration: A comparison with the extended Kalman filter - [논문 링크](https://onlinelibrary.wiley.com/doi/abs/10.1002/navi.421)
Monocular Semantic Mapping Based on 3D Cuboids Tracking - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9401071)
Adaptive Stereo Direct Visual Odometry with Real-Time Loop Closure Detection and Relocalization - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9401469)
DeepAVO: Efficient Pose Refining with Feature Distilling for Deep Visual Odometry - [논문 링크](https://arxiv.org/pdf/2105.09899.pdf)
Vehicle Odometry with Camera-Lidar-IMU Information Fusion and Factor-Graph Optimization - [논문 링크](https://link.springer.com/article/10.1007/s10846-021-01329-x)
RANSIC: Fast and Highly Robust Estimation for Rotation Search and Point Cloud Registration using Invariant Compatibility - [논문 링크](https://arxiv.org/pdf/2104.09133.pdf)
Monocular Semantic Mapping Based on 3D Cuboids Tracking - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9401071)
A sky region segmentation method for outdoor visual-inertial SLAM - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9408771)
Line Flow Based Simultaneous Localization and Mapping - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9393474)
MinkLoc++: Lidar and Monocular Image Fusion for Place Recognition - [논문 링크](https://arxiv.org/pdf/2104.05327.pdf)
Semantic visual SLAM in dynamic environment - [논문 링크](https://link.springer.com/article/10.1007/s10514-021-09979-4)
Simultaneous Video Stabilization and Rolling Shutter Removal - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9411701)
Semi-Direct Monocular SLAM With Three Levels of Parallel Optimizations - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9398863)
Non-uniform Observability for FastMoving Horizon Estimation with application to the SLAM problem - [논문 링크](https://arxiv.org/pdf/2104.12328.pdf)
Radar SLAM: A Robust SLAM System for All Weather Conditions - [논문 링크](https://arxiv.org/pdf/2104.05347.pdf)
Not Only Look But Infer: Multiple Hypothesis Clustering of Data Association Inference for Semantic SLAM - [논문 링크](https://ieeexplore.ieee.org/abstract/document/9410610)
SE-Harris and eSUSAN: Asynchronous Event-Based Corner Detection Using Megapixel Resolution CeleX-V Camera - [논문 링크](https://deepai.org/publication/se-harris-and-esusan-asynchronous-event-based-corner-detection-using-megapixel-resolution-celex-v-camera)
A scalable parallel preconditioned conjugate gradient method for bundle adjustment - [논문 링크](https://link.springer.com/article/10.1007/s10489-021-02349-8)
Self-Supervised Multi-Frame Monocular Scene Flow - [논문 링크](https://arxiv.org/abs/2105.02216)
MonodepthPlus: self-supervised monocular depth estimation using soft-attention and learnable outlier-masking - [논문 링크](https://www.spiedigitallibrary.org/journals/journal-of-electronic-imaging/volume-30/issue-2/023017/MonodepthPlus--self-supervised-monocular-depth-estimation-using-soft-attention/10.1117/1.JEI.30.2.023017.short?SSO=1)