mlsanityData to 7 Oct 2026

MegaSaM: Accurate, Fast, and Robust Structure and Motion from Casual Dynamic Videos

arXiv 2412.04463 ↗First posted 5 Dec 2024cs.CVcited by 224

Not on the map. Why?

Cited by 21 in Diffusion models, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

We present a system that allows for accurate, fast, and robust estimation of camera parameters and depth maps from casual monocular videos of dynamic scenes. Most conventional structure from motion and monocular SLAM techniques assume input videos that feature predominantly static scenes with large amounts of parallax. Such methods tend to produce erroneous estimates in the absence of these conditions. Recent neural network-based approaches attempt to overcome these challenges; however, such methods are either computationally expensive or brittle when run on dynamic videos with uncontrolled camera motion or unknown field of view. We demonstrate the surprising effectiveness of a deep visual SLAM framework: with careful modifications to its training and inference schemes, this system can scale to real-world videos of complex dynamic scenes with unconstrained camera paths, including videos with little camera parallax. Extensive experiments on both synthetic and real videos demonstrate that our system is significantly more accurate and robust at camera pose and depth estimation when compared with prior and concurrent work, with faster or comparable running times. See interactive results on our project page: https://mega-sam.github.io/

Key facts. Cited by 224 papers in the map (data through 7 Oct 2026). Builds on 56 of its 80 references that are in the map. First posted on arXiv 5 Dec 2024.

Builds on

56 of 80 references.

References · 56

  1. 2017Mask R-CNNcited by 7,259
  2. 2022Instant Neural Graphics Primitives with a Multiresolution Hash Encodingcited by 3,698
  3. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,385
  4. 2017What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?cited by 2,335
  5. 2021Vision Transformers for Dense Predictioncited by 1,789
  6. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,526
  7. 2019Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfercited by 1,484
  8. 2024Depth Anything V2cited by 1,385
  9. 2024Depth Anything: Unleashing the Power of Large-Scale Unlabeled Datacited by 1,346
  10. 2015ORB-SLAM: a Versatile and Accurate Monocular SLAM Systemcited by 983
  11. 2024Grounding Image Matching in 3D with MASt3Rcited by 960
  12. 2020ORB-SLAM3: An Accurate Open-Source Library for Visual, Visual-Inertial and Multi-Map SLAMcited by 862
38 more
  1. 2016Unsupervised Monocular Depth Estimation with Left-Right Consistencycited by 819
  2. 2020Neural Scene Flow Fields for Space-Time View Synthesis of Dynamic Scenescited by 750
  3. 2018MegaDepth: Learning Single-View Depth Prediction from Internet Photoscited by 731
  4. 2021DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Camerascited by 673
  5. 2021BARF: Bundle-Adjusting Neural Radiance Fieldscited by 641
  6. 2020TartanAir: A Dataset to Push the Limits of Visual SLAMcited by 494
  7. 2016Direct Sparse Odometrycited by 479
  8. 2024MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motioncited by 431
  9. 2024UniDepth: Universal Monocular Metric Depth Estimationcited by 375
  10. 2022Kubric: A scalable dataset generatorcited by 373
  11. 2023Metric3D: Towards Zero-shot Metric 3D Prediction from A Single Imagecited by 357
  12. 2024Shape of Motion: 4D Reconstruction from a Single Videocited by 247
  13. 2023Visual Geometry Grounded Deep Structure From Motioncited by 225
  14. 2022DynIBaR: Neural Dynamic Image-Based Renderingcited by 224
  15. 2020Consistent Video Depth Estimationcited by 220
  16. 2020Learning to Recover 3D Scene Shape from a Single Imagecited by 214
  17. 2020D3VO: Deep Depth, Deep Pose and Deep Uncertainty for Monocular Visual Odometrycited by 206
  18. 2024DepthCrafter: Generating Consistent Long Depth Sequences for Open-world Videoscited by 202
  19. 2022Monocular Dynamic View Synthesis: A Reality Checkcited by 199
  20. 2023Robust Dynamic Radiance Fieldscited by 193
  21. 2023COLMAP-Free 3D Gaussian Splattingcited by 193
  22. 2020Robust Consistent Video Depth Estimationcited by 188
  23. 2022Deep Patch Visual Odometrycited by 186
  24. 2024MoSca: Dynamic Gaussian Fusion from Casual Videos via 4D Motion Scaffoldscited by 177
  25. 2017DeepVO: Towards End-to-End Visual Odometry with Deep Recurrent Convolutional Neural Networkscited by 173
  26. 2018BA-Net: Dense Bundle Adjustment Networkcited by 173
  27. 2024CAT4D: Create Anything in 4D with Multi-View Video Diffusion Modelscited by 160
  28. 2018CodeSLAM - Learning a Compact, Optimisable Representation for Dense Visual SLAMcited by 151
  29. 2019Learning the Depths of Moving People by Watching Frozen Peoplecited by 120
  30. 2020DeepFactors: Real-Time Probabilistic Dense Monocular SLAMcited by 117
  31. 2023The Surprising Effectiveness of Diffusion Models for Optical Flow and Monocular Depth Estimationcited by 116
  32. 2024Learning Temporally Consistent Video Depth from Video Diffusion Priorscited by 100
  33. 2022ParticleSfM: Exploiting Dense Point Trajectories for Localizing Moving Cameras in the Wildcited by 90
  34. 2022SimpleRecon: 3D Reconstruction Without 3D Convolutionscited by 83
  35. 2024Scene Coordinate Reconstruction: Posing of Image Collections via Incremental Learning of a Relocalizercited by 73
  36. 2023Detector-Free Structure from Motioncited by 71
  37. 2021Consistent Depth of Moving Objects in Videocited by 56
  38. 2023CamP: Camera Preconditioning for Neural Radiance Fieldscited by 55

50 most cited of 56 All 56 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 1; 2025: 109; 2026: 114.

224 papers.

Sort:

By area:

  1. 2025Continuous 3D Perception Model with Persistent Statecited by 527
  2. 2025ReCamMaster: Camera-Controlled Generative Rendering from A Single Videocited by 278
  3. 2025MoGe-2: Accurate Monocular Geometry with Metric Scale and Sharp Detailscited by 250
  4. 2025Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligencecited by 206
  5. 2025ViPE: Video Pose Engine for 3D Geometric Perceptioncited by 162
  6. 2026Advancing Open-source World Modelscited by 144
  7. 2025Video World Models with Long-term Spatial Memorycited by 124
  8. 2025TTT3R: 3D Reconstruction as Test-Time Trainingcited by 122
  9. 2024Stereo4D: Learning How Things Move in 3D from Internet Stereo Videoscited by 89
  10. 2025Easi3R: Estimating Disentangled Motion from DUSt3R Without Trainingcited by 88
  11. 2025St4RTrack: Simultaneous 4D Reconstruction and Tracking in the Worldcited by 88
  12. 2025Geometry Forcing: Marrying Video Diffusion and 3D Representation for Consistent World Modelingcited by 86
38 more
  1. 2025SpatialTrackerV2: 3D Point Tracking Made Easycited by 84
  2. 2025STream3R: Scalable Sequential 3D Reconstruction with Causal Transformercited by 76
  3. 2025Uni3C: Unifying Precisely 3D-Enhanced Camera and Human Motion Controls for Video Generationcited by 75
  4. 2025SpatialVID: A Large-Scale Video Dataset with Spatial Annotationscited by 72
  5. 2025TAPIP3D: Tracking Any Point in Persistent 3D Geometrycited by 67
  6. 2025Large Video Planner Enables Generalizable Robot Controlcited by 67
  7. 2025Geo4D: Leveraging Video Generators for Geometric 4D Scene Reconstructioncited by 66
  8. 2025Scalable Vision-Language-Action Model Pretraining for Robotic Manipulation with Real-Life Human Activity Videoscited by 63
  9. 2025Towards Understanding Camera Motions in Any Videocited by 62
  10. 2025OmniWorld: A Multi-Domain and Multi-Modal Dataset for 4D World Modelingcited by 59
  11. 2025Visual Imitation Enables Contextual Humanoid Controlcited by 57
  12. 2025Evo-0: Vision-Language-Action Model with Implicit Spatial Understandingcited by 55
  13. 2025WildGS-SLAM: Monocular Gaussian Splatting SLAM in Dynamic Environmentscited by 48
  14. 20254DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videoscited by 46
  15. 2025Efficiently Reconstructing Dynamic Scenes One D4RT at a Timecited by 46
  16. 2025CamCloneMaster: Enabling Reference-based Camera Control for Video Generationcited by 44
  17. 2025NovaFlow: Zero-Shot Manipulation via Actionable Flow from Generated Videoscited by 38
  18. 2025Human3R: Everyone Everywhere All at Oncecited by 35
  19. 2026LoGeR: Long-Context Geometric Reconstruction with Hybrid Memorycited by 33
  20. 2026DreamX-World 1.0: A General-Purpose Interactive World Modelcited by 33
  21. 2025Any4D: Unified Feed-Forward Metric 4D Reconstructioncited by 32
  22. 2026LingBot-Map: Geometric Context Transformer for Streaming 3D Reconstructioncited by 32
  23. 2025MoVieS: Motion-Aware 4D Dynamic View Synthesis in One Secondcited by 31
  24. 20254DNeX: Feed-Forward 4D Generative Modeling Made Easycited by 31
  25. 2025Trace Anything: Representing Any Video in 4D via Trajectory Fieldscited by 31
  26. 2025Astra: General Interactive World Model with Autoregressive Denoisingcited by 30
  27. 2025What Is The Best 3D Scene Representation for Robotics? From Geometric to Foundation Modelscited by 29
  28. 2026V-DPM: 4D Video Reconstruction with Dynamic Point Mapscited by 29
  29. 2025Pre-training Auto-regressive Robotic Models with 4D Representationscited by 27
  30. 2025Dynamic Camera Poses and Where to Find Themcited by 26
  31. 2025VGGT4D: Mining Motion Cues in Visual Geometry Transformers for 4D Scene Reconstructioncited by 26
  32. 2025Back on Track: Bundle Adjustment for Dynamic Scene Reconstructioncited by 23
  33. 2025Advances in Feed-Forward 3D Reconstruction and View Synthesis: A Surveycited by 23
  34. 2026VGGT-SLAM 2.0: Real-time Dense Feed-forward Scene Reconstructioncited by 22
  35. 2025Generative Video Motion Editing with 3D Point Trackscited by 21
  36. 2025UniLat3D: Geometry-Appearance Unified Latents for Single-Stage 3D Generationcited by 20
  37. 2026VGGT-Omegacited by 20
  38. 2025iTACO: Interactable Digital Twins of Articulated Objects from Casually Captured RGBD Videoscited by 18

50 most cited of 224 All 224 (JSON)