DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras

By Zachary Teed, Jia Deng. Neural Information Processing Systems, 2021

1,103 citations recorded by the source corpus.

We introduce DROID-SLAM, a new deep learning based SLAM system. DROID-SLAM consists of recurrent iterative updates of camera pose and pixelwise depth through a Dense Bundle Adjustment layer. DROID-SLAM is accurate, achieving large improvements over prior work, and robust, suffering from substantially fewer catastrophic failures. Despite training on monocular video, it can leverage stereo or RGB-D video to achieve improved performance at test time. The URL to our open source code is https://github.com/princeton-vl/DROID-SLAM.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/67515d1f7df144683b059e684da7974e40aeaca1

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Gaussian Splatting SLAM (2023, neutral): “Frame-centric approaches [1, 36] are efficient, as individual frames host local rather than global geometry (e.g. depth maps), and are attractive for long-session SLAM, but if a dense global map is needed, it must be constructed on demand by assembling all of these parts which are not necessarily…”
Co-SLAM: Joint Coordinate and Sparse Parametric Encodings for Neural Real-Time SLAM (2023, background): “Several recent works [11, 29, 30, 38] leverage deep learning to improve the accuracy and robustness of traditional SLAM, and even achieve dense reconstruction with monocular SLAM.”
GS-SLAM: Dense Visual SLAM with 3D Gaussian Splatting (2023, background): “…works integrate deep learning with the traditional geometry framework for more accurate and robust camera tracking and mapping, such as DROID-SLAM [34], CodeSLAM [1], SceneCode [50], and NodeSLAM [31], have made significant advances in the field, achieving more accurate and robust camera tracking…”
Metric3D: Towards Zero-shot Metric 3D Prediction from A Single Image (2023, methodology): “We use our model to predict metric depths, and then naively input them to the Droid-SLAM system as an initial depth.”
SplaTAM: Splat, Track & Map 3D Gaussians for Dense RGB-D SLAM (2023, methodology): “Furthermore, SplaTAM provides better or competitive performance to a feature-based tracking method such as DROID-SLAM [39].”
CoTracker: It is Better to Track Together (2023, background): “Here, we are interested in estimating point correspondences in videos containing dynamic objects and a moving camera.”
Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass (2025, positive): “The latest multi-view geometry pipelines like COLMAP [44] or OrbSLAM2 [30] incorporate nearly 60 years of compounding engineering improvements, but these approaches still catastrophically fail > 40% of the time on static scenes like ETH-3D [52]), which can actually be considered an easy case due to ”
NeRF-SLAM: Real-Time Dense Monocular SLAM with Neural Radiance Fields (2022, methodology): “As our SLAM frontend, we use Droid-SLAM[12], and re-use their pre-trained weights.”
Vox-Fusion: Dense Tracking and Mapping with Voxel-based Neural Implicit Representation (2022, background): “Another successful design is to learn feature matching and scene reconstruction separately [33].”
A Scaling Law for Syn2real Transfer: How Much Is Your Pre-training Effective? (2021, background): “Teed and Deng [59] proposed a simultaneous localization and mapping (SLAM) system that was trained only with synthetic data and demonstrated state-of-the-art performance.”
Gen3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (2025, positive): “If not, we estimate the camera poses using DROID-SLAM [52].”
DUSt3R: Geometric 3D Vision Made Easy (2023, positive): “Recent works on dense visual SLAM from RGB video stream are able to produce high-quality depth maps and camera trajectories [7, 22, 116, 122, 130, 132], but they inherit the traditional limitations of SLAM, e.g . noisy predictions, drifts and out-liers in the pixel correspondences.”
VGGT: Visual Geometry Grounded Transformer (2025, positive): “Recent methods [ 3, 67, 71, 74, 75, 78, 81, 83, 89, 108] explored end-to-end differentiable SfM, where VG-GSfM [83] started to outperform traditional algorithms on challenging phototourism scenarios.”
MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion (2024, background): “Given a set of 2D images, structure from motion (SfM) (Sch¨onberger & Frahm, 2016; Teed & Deng, 2018; Tang & Tan, 2018) or visual SLAM (Teed & Deng, 2021; Mur-Artal et al., 2015; Mur-Artal & Tard´os, 2017; Engel et al., 2014; Newcombe et al., 2011) estimate 3D structure of a scene while also…”
Metric3D v2: A Versatile Monocular Geometric Foundation Model for Zero-Shot Metric Depth and Surface Normal Estimation (2024, positive): “Fortunately, we observe that robust metric depth models are scalable geometric learners, containing abundant information for normal estimation.”
NICE-SLAM: Neural Implicit Scalable Encoding for SLAM (2021, methodology): “DROID-SLAM [52] uses regressed optical flow to define geometrical residuals for its refinement.”
Continuous 3D Perception Model with Persistent State (2025, neutral): “Monocular SLAM pipelines [22, 23, 78, 105] recover ego-motion and 3D point clouds in real-time from video sequences but typically require known camera intrinsics.”
Adaptive VIO: Deep Visual-Inertial Odometry with Online Continual Learning (2024, methodology): “DROID-SLAM [33] and DPVO [35] combine iterative visual correspondence updates with dif ferentiable bundle adjustment, demonstrating excellent per formance across multiple datasets. iSLAM [12] combines learning-based VO with graph optimization, further enhanc ing performance through self-supervised…”