Awesome Learning-based Odometry

This repository contains a curated list of resources addressing Learning-based visual odometry (including both image and event camera), visual-inertial odometry, inertial odometry, LiDAR odometry, Semantic SLAM, NeRF SLAM, etc.

If you find some ignored papers, feel free to create pull requests, or open issues.

Contributions in any form to make this list more comprehensive are welcome.

If you find this repository useful, a simple star should be the best affirmation. 😊

Feel free to share this list with others!

Overview

• Learning-based VO
• Learning-based VIO
• Learning-based Inertial Odometry
• Learning-based LiDAR Odometry
• Semantic SLAM
• NeRF SLAM
• Other Related Resource

Learning-based VO

Year	Venue	Paper Title	Repository	Note
2026	CVPR	DROID-W: DROID-SLAM in the Wild		website 不去“显式建模动态物体”，而是通过不确定性建模（uncertainty）自动忽略不可靠区域
2026	CVPR	SCE-SLAM: Scale-Consistent Monocular SLAM via Scene Coordinate Embeddings	---	基于DPVO的双分支单目SLAM，用scene coordinate embeddings在canonical scale下维持跨窗口尺度一致性（非预测绝对metric尺度）：CNN加DINOv3提特征，按SuperPoint每帧采样patch；flow分支做像素级跟踪，scene coordinate分支以patch级表征累积尺度记忆；Geometry-Guided Scale Propagation用几何调制注意力从空间邻近的历史参考patch传播尺度，经GRU与帧内聚合更新；Scene Coordinate Bundle Adjustment将embedding解码的三维坐标先验作为世界系约束，与光流残差交替优化，把漂移拉回初始化尺度
2026	arXiv	Marginalized Bundle Adjustment: Multi-View Camera Pose from Monocular Depth Estimates	---	website 首个将通用单目深度估计（MDE）模型集成到不同尺度的运动恢复结构（SfM）和相机重定位任务中的框架
2026	RAL	360DVO: Deep Visual Odometry for Monocular 360-Degree Camera		website 基于深度学习的单目全向视觉里程计框架,畸变感知球面特征提取器(SphereResNet)+全向可微分束调整 (ODBA) 模块
2026	arXiv	RADIO-ViPE: Online Tightly Coupled Multi-Modal Fusion for Open-Vocabulary Semantic SLAM in Dynamic Environments		website RADIO-ViPE = 基于DROID-SLAM的架构，在 ViPE 上做 视觉–语言–几何紧耦合 的在线 Semantic SLAM；输入只要未标定单目 RGB，就能建图、定位，并支持开放词汇 3D 查询。GeoCalib 进行相机的初始化（从均匀采样的帧中引导出相机内参，无需标定目标或已知相机模型） + RADSeg/RADIO 稠密特征 + 单目深度；在 BA 中联合光流几何项与RADIO embedding 相似项；时序自适应鲁棒核处理动态/准静态物体；SigLIP 做开放词汇 3D grounding；
2025	arXiv	Vipe: Video pose engine for 3d geometric perception		website 英伟达开源的，从视频流中估算相机pose与深度；将BA（类似 DROID-SLAM的稠密光流约束）与learning深度估计相结合：语义分割移动物体+DROID-SLAM构建BA约束+单目深度估计网络（Metric3dv2）； RTX 5090，640*480分辨率，3-5HZ
2026	AAAI	FoundationSLAM: Unleashing the Power of Depth Foundation Models for End-to-End Dense Visual SLAM	---	通过基础深度模型提供的指导，将光流估计/flow estimation与几何推理相结合； Hybrid Flow Network输出几何约束的correspondences，进而保证深度与pose估计的一致性； Bi-Consistent Bundle Adjustment Layer对多视角约束下的关键帧pose和深度进行优化；Reliability-Aware Refinement mechanism通过区分可靠与不可靠的区域实现flow update
2025	arXiv	KM-ViPE: Online Tightly Coupled Vision-Language-Geometry Fusion for Open-Vocabulary Semantic SLAM		输入RGB数据，经GeoCalib估计相机内参、DINO提取视觉特征、自适应鲁棒kernel处理动态区域，再通过global BA优化位姿和深度，最后融合CLIP语言编码器实现开放词汇查询，标注“视觉-几何-语言全链路融合，实现“几何建图+语义理解+动态适应”
2025	TRO	Continual Learning of Regions for Efficient Robot Localization on Large Maps		---
2025	arXiv	Policies over Poses: Reinforcement Learning based Distributed Pose-Graph Optimization for Multi-Robot SLAM		---
2025	arXiv	Real-Time Indoor Object SLAM with LLM-Enhanced Priors	---	---
2025	TIM	A Robust and Accurate Stereo SLAM Based on Learned Feature Extraction and Matching	---	---
2025	DAGM German Conference on Pattern Recognition	CoProU-VO: Combining Projected Uncertainty for End-to-End Unsupervised Monocular Visual Odometry		website
2025	RAL	Occlusion-Aware Monocular Visual Odometry for Robust Trajectory Tracking	---	---
2025	RAL	DINO-VO: A Feature-Based Visual Odometry Leveraging a Visual Foundation Model	---	---
2025	IEEE International Symposium on Circuits and Systems	LoSeVO: Local Sequence Constraints for Deep Visual Odometry	---	---
2025	arXiv	VOCAL: Visual Odometry via ContrAstive Learning	---	---
2025	RAL	Online Adaptive Keypoint Extraction for Visual Odometry Across Different Scenes	---	Reinforcement Learning
2025	ICRA	MAC-VO: Metrics-aware Covariance for Learning-based Stereo Visual Odometry		website
2025	arXiv	Large-scale visual SLAM for in-the-wild videos	---	---
2025	CVPR	MegaSaM: Accurate, Fast, and Robust Structure and Motion from Casual Dynamic Videos		website 从含运动物体、视差可能很小、焦距可能未知的随意单目视频中估计相机位姿、焦距和时序一致深度。在 DROID-SLAM 的可微光束法平差上：另预测物体运动概率图，用来改变动态区域在平差中的权重；先在静态视频上学自运动光流，再冻结该网络、只在动态视频上训练运动模块。推理用 Depth Anything 的相对视差和 UniDepth 的度量尺度、焦距做初始化；视差充足时主要靠重投影，视差不可观测（如近纯旋转）时才加入单目深度正则，焦距不可观测时则关闭焦距优化。位姿固定后，直接优化视差与不确定性得到更高分辨率的一致深度，而不对深度网络做逐视频微调。
2025	CVPR	Scene-agnostic Pose Regression for Visual Localization		website Mamba-based
2025	arXiv	Image as an IMU: Estimating Camera Motion from a Single Motion-Blurred Image	---	Velometer
2024	CVPR	From variance to veracity: Unbundling and mitigating gradient variance in differentiable bundle adjustment layers		---
2024	Complex & Intelligent Systems	Lp-slam: language-perceptive RGB-D SLAM framework exploiting large language model		---
2024	TIV	Dh-ptam: a deep hybrid stereo events-frames parallel tracking and mapping system		Superpoint+stereo ptam
2024	RAL	Salient sparse visual odometry with pose-only supervision	---	---
2024	RAL	CodedVO: Coded Visual Odometry		website
2024	CVPR	Leap-vo: Long-term effective any point tracking for visual odometry		website
2024	RAL	Efficient Camera Exposure Control for Visual Odometry via Deep Reinforcement Learning		---
2024	ECCV	Reinforcement learning meets visual odometry		Blog
2024	IROS	Deep Visual Odometry with Events and Frames		RAMP-VO
2024	3DV	Deep event visual odometry		---
2024	CVPR	Multi-Session SLAM with Differentiable Wide-Baseline Pose Optimization		Multi VO
2024	ECCV	Deep patch visual slam		---
2024	NIPS	Deep patch visual odometry		---
2023	CVPR	Pvo: Panoptic visual odometry		website
2023	ICRA	Dytanvo: Joint refinement of visual odometry and motion segmentation in dynamic environments		---
2022	IEEE/SICE International Symposium on System Integration	Maskvo: Self-supervised visual odometry with a learnable dynamic mask	---	---
2022	Sensor	Raum-vo: Rotational adjusted unsupervised monocular visual odometry	---	---
2022	Neurocomputing	DeepAVO: Efficient pose refining with feature distilling for deep Visual Odometry	---	---
2021	CoRL	Tartanvo: A generalizable learning-based vo		---
2021	NIPS	DROID-SLAM: Deep Visual SLAM for Monocular,Stereo, and RGB-D Cameras		---
2021	CVPR	Generalizing to the open world: Deep visual odometry with online adaptation	---	---
2021	arXiv	DF-VO: What should be learnt for visual odometry?		---
2021	ICRA	CNN-based ego-motion estimation for fast MAV maneuvers		---
2020	ICRA	Visual odometry revisited: What should be learnt?		---
2020	CVPR	Towards better generalization: Joint depth-pose learning without posenet		---
2020	CVPR	Diffposenet: Direct differentiable camera pose estimation	---	---
2020	CVPR	D3vo: Deep depth, deep pose and deep uncertainty for monocular visual odometry	---	website
2021	IJCV	Unsupervised scale-consistent depth learning from video		---
2022	RAL	SimVODIS++: Neural semantic visual odometry in dynamic environments	---	---
2020	TPAMI	Simvodis: Simultaneous visual odometry, object detection, and instance segmentation		---
2019	NIPS	Unsupervised scale-consistent depth and ego-motion learning from monocular video		---
2019	ICRA	Ganvo: Unsupervised deep monocular visual odometry and depth estimation with generative adversarial networks	---	---
2019	CVPR	Recurrent neural network for (un-) supervised learning of monocular video visual odometry and depth	-	-
2019	ICRA	Pose graph optimization for unsupervised monocular visual odometry	---	---
2018	IJRR	End-to-end, sequence-to-sequence probabilistic visual odometry through deep neural networks	---	---
2018	CVPR	Unsupervised learning of monocular depth estimation and visual odometry with deep feature reconstruction		---
2018	ICRA	Undeepvo: Monocular visual odometry through unsupervised deep learning	-	website
2017	ICRA	Deepvo: Towards end-to-end visual odometry with deep recurrent convolutional neural networks	-	website
2017	CVPR	Unsupervised learning of depth and ego-motion from video		---
2015	RAL	Exploring representation learning with cnns for frame-to-frame ego-motion estimation	---	---

Learning-based VIO

• Survey for Awesome Learning-based VO and VIO Blog

Year	Venue	Paper Title	Repository	Note
2025	TRO	OKVIS2-X: Open Keyframe-based Visual-Inertial SLAM Configurable with Dense Depth or LiDAR, and GNSS		---
2025	arXiv	SuperPoint-SLAM3: Augmenting ORB-SLAM3 with Deep Features, Adaptive NMS, and Learning-Based Loop Closure		---
2025	arXiv	SuperEvent: Cross-Modal Learning of Event-based Keypoint Detection		website
2025	arXiv	SuperEIO: Self-Supervised Event Feature Learning for Event Inertial Odometry		website
2025	Robotics and Autonomous Systems	CUAHN-VIO: Content-and-uncertainty-aware homography network for visual-inertial odometry		---
2025	TRO	Airslam: An efficient and illumination-robust point-line visual slam system		---
2024	arXiv	SF-Loc: A Visual Mapping and Geo-Localization System based on Sparse Visual Structure Frames		test
2024	arXiv	DEIO: Deep Event Inertial Odometry		website
2024	CVPR	Adaptive vio: Deep visual-inertial odometry with online continual learning	---	---
2024	RAL	DBA-Fusion: Tightly Integrating Deep Dense Visual Bundle Adjustment with Multiple Sensors for Large-Scale Localization and Mapping		---
2024	ICRA	DVI-SLAM: A dual visual inertial SLAM network	---	---
2023	ICRA	Bamf-slam: bundle adjusted multi-fisheye visual-inertial slam using recurrent field transforms	---	---
2022	ECCV	Efficient deep visual and inertial odometry with adaptive visual modality selection		---
2022	IEEE/ASME International Conference on Advanced Intelligent Mechatronics	A self-supervised, differentiable Kalman filter for uncertainty-aware visual-inertial odometry	---	---
2022	Neural Networks	SelfVIO: Self-supervised deep monocular Visual–Inertial Odometry and depth estimation	---	---
2021	International Conference on International Joint Conferences on Artificial Intelligence	Unsupervised monocular visual-inertial odometry network		---
2021	IEEE International Conference on Acoustics, Speech and Signal Processing	Atvio: Attention guided visual-inertial odometry	---	---
2019	TPAMI	Unsupervised deep visual-inertial odometry with online error correction for RGB-D imagery	---	---
2019	CVPR	Selective sensor fusion for neural visual-inertial odometry	---	---
2019	IEEE Global Communications Conference	LightVO: Lightweight inertial-assisted monocular visual odometry with dense neural networks	---	---
2019	IROS	Deepvio: Self-supervised deep learning of monocular visual inertial odometry using 3d geometric constraints	---	---
2017	AAAI	Vinet: Visual-inertial odometry as a sequence-to-sequence learning problem		non-official implementation

Learning-based Inertial Odometry

• Survey for Deep IMU-Bias Inference Blog

Year	Venue	Paper Title	Repository	Note
2026	RAL	X-IONet: Cross-Platform Inertial Odometry Network for Pedestrian and Legged Robot	---	跨平台纯IMU惯性里程计（行人+四足）：先用EKF姿态将IMU旋到重力对齐坐标系；再以1D-CNN对窗口做平台二分类（行人/四足），按规则路由到对应平台专家网络；各专家采用受Crossformer启发的双阶段注意力编解码器——时间自注意力建模长程依赖、维度自注意力建模gyro/acc轴间耦合，并分层多尺度特征融合——回归3D位移及其协方差；以Huber-Gaussian损失联合优化位移与不确定性，再将网络量测送入EKF做状态更新以抑制漂移；在RoNIN、GrandTour与自采Go2上验证跨平台有效性
2025	arXiv	AirIO: Learning Inertial Odometry with Enhanced IMU Feature Observability		website Test
2024	TIV	Deep learning for inertial positioning: A survey	---	---
2023	arXiv	End-to-end deep learning framework for real-time inertial attitude estimation using 6dof imu	---	---
2023	arXiv	AirIMU: Learning uncertainty propagation for inertial odometry		website Test
2023	RAL	Learned inertial odometry for autonomous drone racing		---
2022	ICRA	Improved state propagation through AI-based pre-processing and down-sampling of high-speed inertial data	---	---
2022	RAL	Deep IMU Bias Inference for Robust Visual-Inertial Odometry with Factor Graphs	---	---
2021	ICRA	IMU Data Processing For Inertial Aided Navigation:A Recurrent Neural Network Based Approach	---	---
2020	RAL	Tlio: Tight learned inertial odometry		website
2020	TIV	AI-IMU dead-reckoning		---

Learning-based LiDAR Odometry

• Survey for Awesome Learning-based LiDAR Odometry Blog
• LiDAR-based 3DGS Paper List

Year	Venue	Paper Title	Repository	Note
2025	arXiv	MA-SLAM: Active SLAM in Large-Scale Unknown Environment using Map Aware Deep Reinforcement Learning	---	---
2025	IEEE Sensor Journal	A Generative Hierarchical Optimization Framework for LiDAR Odometry Using Conditional Diffusion Models	---	---
2025	CVPR	DiffLO: Semantic-Aware LiDAR Odometry with Diffusion-Based Refinement		---
2025	arXiv	LIR-LIVO: A Lightweight, Robust LiDAR/Vision/Inertial Odometry with Illumination-Resilient Deep Features		Fast-LIVO+AirSLAM
2024	arXiv	LiDAR Inertial Odometry And Mapping Using Learned Registration-Relevant Features		learning-based select points
2024	IEEE 100th Vehicular Technology Conference	LiDAR-OdomNet: LiDAR Odometry Network Using Feature Fusion Based on Attention		---
2024	IROS	LiDAR-Visual-Inertial Tightly-coupled Odometry with Adaptive Learnable Fusion Weights	---	F-LOAM+VINS-Mono
2024	TRO	PIN-SLAM: LiDAR SLAM Using a Point-Based Implicit Neural Representation for Achieving Global Map Consistency		---
2024	AAAI	DeepPointMap: Advancing LiDAR SLAM with Unified Neural Descriptors		---
2023	ICCV	NeRF-LOAM: Neural Implicit Representation for Large-Scale Incremental LiDAR Odometry and Mapping		---
2023	ICCV	DELO: Deep Evidential LiDAR Odometry using Partial Optimal Transport		website
2023	AAAI	Translo: A window-based masked point transformer framework for large-scale lidar odometry		---
2023	RAL	LONER: LiDAR Only Neural Representations for Real-Time SLAM		---
2023	TIV	HPPLO-Net: Unsupervised LiDAR Odometry Using a Hierarchical Point-to-Plane Solver		---
2022	TPAMI	Efficient 3D Deep LiDAR Odometry		---
2021	ICRA	Self-supervised learning of lidar odometry for robotic applications		---
2021	CVPR	PWCLO-Net: Deep LiDAR Odometry in 3D Point Clouds Using Hierarchical Embedding Mask Optimization		---
2021	ISPRS	Deeplio: deep lidar inertial sensor fusion for odometry estimation		---
2020	ACM international conference on multimedia	Lodonet: A deep neural network with 2d keypoint matching for 3d lidar odometry estimation	---	---
2020	ICRA	Unsupervised geometry-aware deep lidar odometry	---	website
2020	IROS	DMLO: Deep Matching LiDAR Odometry	---	---
2019	IROS	Deeppco: End-to-end point cloud odometry through deep parallel neural network	---	---
2019	CVPR	Lo-net: Deep real-time lidar odometry	---	---
2019	CVPR	L3-net: Towards learning based lidar localization for autonomous driving	---	---
2018	IEEE International Conference on Autonomous Robot Systems and Competitions	CNN for IMU assisted odometry estimation using velodyne LiDAR	---	---
2016	RSS workshop	Deep learning for laser based odometry estimation	---	---

Semantic SLAM

Year	Venue	Paper Title	Repository	Note
2026	RAL	SLIM-VDB: A Real-Time 3D Probabilistic Semantic Mapping Framework		利用 OpenVDB 数据结构(用于高效的体素数据存储)，并引入贝叶斯语义融合更新，以实现针对闭集或开集语义的实时、内存高效的语义建图；对于输入的RGB-D或者点云，先通过SAM来做语义分割，然后用TSDF进行表面重建，用统一贝叶斯进行语义融合，对应于开集或闭集语义的截断符号距离值、权重和语义参数存储在OpenVDB数据结构的体素中。
2025	TRO	RAZER: Robust Accelerated Zero-Shot 3D Open-Vocabulary Panoptic Reconstruction with Spatio-Temporal Aggregation	---	website 构建开放词汇语义地图,FC-CLIP+ConvNeXt,支持560类物体+1306个文本类别

NeRF SLAM

• Survey for NeRF-based SLAM：Blog

Year	Venue	Paper Title	Repository	Note
2025	TPAMI	SNI-SLAM++: Tightly-Coupled Semantic Neural Implicit SLAM	---	---
2025	CVPR	MNE-SLAM: Multi-Agent Neural SLAM for Mobile Robots		---
2025	arXiv	MCN-SLAM: Multi-Agent Collaborative Neural SLAM with Hybrid Implicit Neural Scene Representation		website
2025	arXiv	LRSLAM: Low-rank Representation of Signed Distance Fields in Dense Visual SLAM System	---	---
2025	arXiv	Joint Optimization of Neural Radiance Fields and Continuous Camera Motion from a Monocular Video		---
2024	RAL	N3-Mapping: Normal Guided Neural Non-Projective Signed Distance Fields for Large-scale 3D Mapping		---
2024	TRO	PIN-SLAM: LiDAR SLAM Using a Point-Based Implicit Neural Representation for Achieving Global Map Consistency		---
2024	AAAI	DeepPointMap: Advancing LiDAR SLAM with Unified Neural Descriptors		---
2024	ICRA	Towards Large-Scale Incremental Dense Mapping using Robot-centric Implicit Neural Representation		---
2024	CVPR	3D LiDAR Mapping in Dynamic Environments using a 4D Implicit Neural Representation		---
2023	ICRA	SHINE-Mapping: Large-Scale 3D Mapping Using Sparse Hierarchical Implicit NEural Representations		---
2023	RAL	LONER: LiDAR Only Neural Representations for Real-Time SLAM		---
2023	ICCV	NeRF-LOAM: Neural Implicit Representation for Large-Scale Incremental LiDAR Odometry and Mapping		---

Other Related Resource

• Survey for Transformer-based SLAM：Paper List
• Survey for Diffusion-based SLAM：Paper List
• Survey for 3DGS-based SLAM: Paper List
• Survey for Dynamic SLAM: Blog
• Deep learning for image matching lecture slides

• Some related papers for learning-based VO, VIO, such as datasets, depth estimation, correspondence learning, etc.

Year	Venue	Paper Title	Repository	Note
2025	arXiv	SupeRANSAC: One RANSAC to Rule Them All		---
2025	ICRA	LiftFeat: 3D Geometry-Aware Local Feature Matching		---
2025	arXiv	EdgePoint2: Compact Descriptors for Superior Efficiency and Accuracy		---
2025	arXiv	To Match or Not to Match: Revisiting Image Matching for Reliable Visual Place Recognition	---	Matching
2025	CVPR	Learning Affine Correspondences by Integrating Geometric Constraints		Matching/Correspondences
2025	CVPR	MP-SfM: Monocular Surface Priors for Robust Structure-from-Motion		---
2025	arXiv	Selecting and Pruning: A Differentiable Causal Sequentialized State-Space Model for Two-View Correspondence Learning	---	Mamba two-view correspondence
2024	CVPR	DeDoDe v2: Analyzing and Improving the DeDoDe Keypoint Detector		DeDoDe v2 Pytorch Inference
2024	3DV	DeDoDe: Detect, don't describe—Describe, don't detect for local feature matching		detectors and descriptors jointly
2024	ECCV	Stereoglue: Robust estimation with single-point solvers		stereo superglue
2024	TPAMI	Metric3d v2: A versatile monocular geometric foundation model for zero-shot metric depth and surface normal estimation		website depth estimation
2022	NIPS	Theseus: A library for differentiable nonlinear optimization		website
2020	CVPR	Superglue: Learning feature matching with graph neural networks		---
2020	IROS	Tartanair: A dataset to push the limits of visual slam		website
2021	3DV	RAFT-Stereo: Multilevel Recurrent Field Transforms for Stereo Matching		---
2020	ECCV	Raft: Recurrent all-pairs field transforms for optical flow		---
2019	CVPR	Projective manifold gradient layer for deep rotation regression		Better learn rotations
2018	ECCV	Mvsnet: Depth inference for unstructured multi-view stereo	---	---
2018	CVPR	Superpoint: Self-supervised interest point detection and description		pytorch version event superpoint
2016	CVPR	Structure-from-motion revisited		COLMAP
2015	ICCV	Flownet: Learning optical flow with convolutional networks	---	---

• Some related papers for learning-based LiDAR odometry, such as point cloud registration, etc.

Year	Venue	Paper Title	Repository	Note
2025	ICCV	TurboReg: TurboClique for Robust and Efficient Point Cloud Registration		---
2024	CVPR	Point transformer v3: Simpler faster stronger		---
2024	NeurIPS	A Consistency-Aware Spot-Guided Transformer for Versatile and Hierarchical Point Cloud Registration		---
2020	CVPR	3dregnet: A deep neural network for 3d point registration		---
2020	CVPR	P2b: Point-to-box network for 3d object tracking in point clouds	---	---
2019	CVPR	Pointconv: Deep convolutional networks on 3d point clouds	---	---
2019	ICCV	Meteornet: Deep learning on dynamic 3d point cloud sequences	---	---
2019	ICCV	Deep closest point: Learning representations for point cloud registration	---	---
2019	CVPR	Pointnetlk: Robust & efficient point cloud registration using pointnet		---
2019	CVPR	3D local features for direct pairwise registration	---	---
2019	ICCV	Kpconv: Flexible and deformable convolution for point clouds		---
2019	ICCV	Deepvcp: An end-to-end deep neural network for point cloud registration	---	---
2019	CVPR	Flownet3d: Learning scene flow in 3d point clouds	---	---
2019	IROS	Rangenet++: Fast and accurate lidar semantic segmentation	---	---
2019	ACM Transactions on Graphics	Dynamic graph cnn for learning on point clouds	---	---
2018	ECCV	3dfeat-net: Weakly supervised local 3d features for point cloud registration		---
2017	NIPS	Pointnet++: Deep hierarchical feature learning on point sets in a metric space	---	---
2017	CVPR	Pointnet: Deep learning on point sets for 3d classification and segmentation	---	---
2015	IROS	Voxnet: A 3d convolutional neural network for real-time object recognition	---	---