CVPR 2021 SLAM 相关论文汇总

作者丨wanghy@知乎

来源丨https://zhuanlan.zhihu.com/p/382419307

编辑丨3D视觉工坊

CVPR 2021 全部论文已经放出，网址：https://openaccess.thecvf.com/CVPR2021?day=all。特总结了与SLAM方向相关的论文，供SLAMer参考。

点云配准：

·RPSRNet: End-to-End Trainable Rigid Point Set Registration Network using Barnes-Hut 2 D -Tree Representation

·SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration

·PointDSC: Robust Point Cloud Registration using Deep Spatial Consistency

·ReAgent: Point Cloud Registration using Imitation and Reinforcement Learning

·Robust Point Cloud Registration Framework Based on Deep Graph Matching

·PREDATOR : Registration of 3D Point Clouds with Low Overlap

·DeepI2P: Image-to-Point Cloud Registration via Deep Classification

点云压缩：

·VoxelContext-Net: An Octree based Framework for Point Cloud Compression

点云上采样&&点云补全：

·Point Cloud Upsampling via Disentangled Refinement

·PU-GCN: Point Cloud Upsampling using Graph Convolutional Networks

·PMP-Net: Point Cloud Completion by Learning Multi-step Point Moving Paths

·View-Guided Point Cloud Completion

·Style-based Point Generator with Adversarial Rendering for Point Cloud Completion

·Cycle4Completion: Unpaired Point Cloud Completion using Cycle Transformation with Missing Region Coding

·Variational Relational Point Completion Network

·Unsupervised 3D Shape Completion through GAN Inversion

定位与建图：

·Privacy Preserving Localization and Mapping from Uncalibrated Cameras

·Differentiable SLAM-net: Learning Particle SLAM for Visual Navigation

·Large-scale Localization Datasets in Crowded Indoor Spaces

·Rotation-Only Bundle Adjustment

·Learning Camera Localization via Dense Scene Matching

·Uncertainty-Aware Camera Pose Estimation from Points and Lines

·DSC-PoseNet: Learning 6DoF Object Pose Estimation via Dual-scale Consistency

·VS-Net: Voting with Segmentation for Visual Localization

·Robust Neural Routing Through Space Partitions for Camera Relocalization in Dynamic Indoor Environments

·Back to the Feature: Learning Robust Camera Localization from Pixels to Pose

里程计：

·Globally Optimal Relative Pose Estimation with Gravity Prior

·Spatiotemporal Registration for Event-based Visual Odometry

·PWCLO-Net: Deep LiDAR Odometry in 3D Point Clouds Using Hierarchical Embedding Mask Optimization

·Generalizing to the Open World: Deep Visual Odometry with Online Adaptation

AR：

·Stay Positive: Non-Negative Image Synthesis for Augmented Reality

·HDR Environment Map Estimation for Real-Time Augmented Reality video

·VITON-HD: High-Resolution Virtual Try-On via Misalignment-Aware Normalization

·Self-Supervised Collision Handling via Generative 3D Garment Models for Virtual Try-On

·ANR: Articulated Neural Rendering for Virtual Avatars

·Toward Accurate and Realistic Outfits Visualization with Attention to Details

闭环&&场景识别：

·OBoW: Online Bag-of-Visual-Words Generation for Self-Supervised Learning

·Patch-NetVLAD: Multi-Scale Fusion of Locally-Global Descriptors for Place Recognition

·SOE-Net: A Self-Attention and Orientation Encoding Network for Point Cloud based Place Recognition

标定：

·Wide-Baseline Multi-Camera Calibration using Person Re-Identification

SFM:

·Efficient Initial Pose-graph Generation for Global SfM

·Deep Two-View Structure-from-Motion Revisited

光流：

·Learning Optical Flow from a Few Matches

·UPFlow: Upsampling Pyramid for Unsupervised Optical Flow Learning

·AutoFlow: Learning a Better Training Set for Optical Flow

·Learning optical flow from still images

场景流：

·Weakly Supervised Learning of Rigid 3D Scene Flow

·Self-Supervised Multi-Frame Monocular Scene Flow

·FlowStep3D: Model Unrolling for Self-Supervised Scene Flow Estimation

·HCRF-Flow: Scene Flow from Point Clouds with Continuous High-order CRFs and Position-aware Flow Embedding

·Self-Point-Flow: Self-Supervised Scene Flow Estimation from Point Clouds with Optimal Transport and Random Walk

Reconstruction:

·Learning to Recover 3D Scene Shape from a Single Image（Finalist）

·NeuralRecon: Real-Time Coherent 3D Reconstruction from Monocular Video （Finalist）

·Neural Deformation Graphs for Globally-consistent Non-rigid Reconstruction

·Multi-view 3D Reconstruction of a Texture-less Smooth Surface of Unknown Generic Reflectance

·Square Root Bundle Adjustment for Large-Scale Reconstruction

·Self-Supervised 3D Mesh Reconstruction from Single Images

·DI-Fusion: Online Implicit 3D Reconstruction with Deep Priors

·Learning monocular 3D reconstruction of articulated categories from motion

·Deep Implicit Moving Least-Squares Functions for 3D Reconstruction

·Plan2Scene: Converting Floorplans to 3D Scenes

·Deep Optimized Priors for 3D Shape Modeling and Reconstruction

depth:

·AdaBins: Depth Estimation Using Adaptive Bins

·Feature-Level Collaboration: Joint Unsupervised Learning of Optical Flow, Stereo Depth and Camera Motion

·Depth from Camera Motion and Object Detection

·Sparse Auxiliary Networks for Unified Monocular Depth Prediction and Completion

·Towards Fast and Accurate Real-World Depth Super Resolution: Benchmark Dataset and Baseline

·Three Ways to Improve Semantic Segmentation with Self-Supervised Depth Estimation

·Learnable Sparse Signal Superdensity for Guided Depth Estimation

·Depth Completion with Twin Surface Extrapolation at Occlusion boundaries

·Differentiable Diffusion for Dense Depth Estimation from Multi-view Images

·Robust Consistent Video Depth Estimation

·Multi-view Depth Estimation using Epipolar Spatio-Temporal Networks

·Radar-Camera Pixel Depth Association for Depth Completion

·Boosting Monocular Depth Estimation Models to High-Resolution via Content-Adaptive Multi-Resolution Merging

·Dual Pixel Exploration: Simultaneous Depth Estimation and Image Restoration

·Beyond Image to Depth: Improving Depth Prediction using Echoes

·SliceNet: deep dense depth estimation from a single indoor panorama using a slice-based representation

·Single Image Depth Prediction with Wavelet Decomposition

·Self-supervised Learning of Depth Inference for Multi-view Stereo

·Camera Pose Matters: Improving Depth Prediction by Mitigating Pose Distribution Bias

本文仅做学术分享，如有侵权，请联系删文。

下载1

在「3D视觉工坊」公众号后台回复：3D视觉，即可下载 3D视觉相关资料干货，涉及相机标定、三维重建、立体视觉、SLAM、深度学习、点云后处理、多视图几何等方向。

下载2

在「3D视觉工坊」公众号后台回复：3D视觉github资源汇总，即可下载包括结构光、标定源码、缺陷检测源码、深度估计与深度补全源码、点云处理相关源码、立体匹配源码、单目、双目3D检测、基于点云的3D检测、6D姿态估计源码汇总等。

下载3

在「3D视觉工坊」公众号后台回复：相机标定，即可下载独家相机标定学习课件与视频网址；后台回复：立体匹配，即可下载独家立体匹配学习课件与视频网址。

重磅！3DCVer-学术论文写作投稿 交流群已成立

扫码添加小助手微信，可申请加入3D视觉工坊-学术论文写作与投稿 微信交流群，旨在交流顶会、顶刊、SCI、EI等写作与投稿事宜。

同时也可申请加入我们的细分方向交流群，目前主要有3D视觉、CV&深度学习、SLAM、三维重建、点云后处理、自动驾驶、多传感器融合、CV入门、三维测量、VR/AR、3D人脸识别、医疗影像、缺陷检测、行人重识别、目标跟踪、视觉产品落地、视觉竞赛、车牌识别、硬件选型、学术交流、求职交流、ORB-SLAM系列源码交流、深度估计等微信群。

一定要备注：研究方向+学校/公司+昵称，例如：”3D视觉 + 上海交大 + 静静“。请按照格式备注，可快速被通过且邀请进群。原创投稿也请联系。

▲长按加微信群或投稿

▲长按关注公众号

3D视觉从入门到精通知识星球：针对3D视觉领域的视频课程（三维重建系列、三维点云系列、结构光系列、手眼标定、相机标定、orb-slam3等视频课程）、知识点汇总、入门进阶学习路线、最新paper分享、疑问解答五个方面进行深耕，更有各类大厂的算法工程人员进行技术指导。与此同时，星球将联合知名企业发布3D视觉相关算法开发岗位以及项目对接信息，打造成集技术与就业为一体的铁杆粉丝聚集区，近2000星球成员为创造更好的AI世界共同进步，知识星球入口：

学习3D视觉核心技术，扫描查看介绍，3天内无条件退款

 圈里有高质量教程资料、可答疑解惑、助你高效解决问题

觉得有用，麻烦给个赞和在看~