mlsanityData to 14 Oct 2026

MV-DUSt3R+: Single-Stage Scene Reconstruction from Sparse Views In 2 Seconds

arXiv 2412.06974 ↗First posted 9 Dec 2024cs.CV cs.AIcited by 141

Not on the map. Why?

Abstract (arXiv)

Recent sparse multi-view scene reconstruction advances like DUSt3R and MASt3R no longer require camera calibration and camera pose estimation. However, they only process a pair of views at a time to infer pixel-aligned pointmaps. When dealing with more than two views, a combinatorial number of error prone pairwise reconstructions are usually followed by an expensive global optimization, which often fails to rectify the pairwise reconstruction errors. To handle more views, reduce errors, and improve inference time, we propose the fast single-stage feed-forward network MV-DUSt3R. At its core are multi-view decoder blocks which exchange information across any number of views while considering one reference view. To make our method robust to reference view selection, we further propose MV-DUSt3R+, which employs cross-reference-view blocks to fuse information across different reference view choices. To further enable novel view synthesis, we extend both by adding and jointly training Gaussian splatting heads. Experiments on multi-view stereo reconstruction, multi-view pose estimation, and novel view synthesis confirm that our methods improve significantly upon prior art. Code will be released.

Key facts. Cited by 141 papers in the map (data through 14 Oct 2026). Builds on 68 of its 99 references that are in the map. First posted on arXiv 9 Dec 2024.

Builds on

68 of 99 references.

References · 68

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  3. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  4. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  5. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,226
  6. 2017ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenescited by 3,117
  7. 2019DeepSDF: Learning Continuous Signed Distance Functions for Shape Representationcited by 2,689
  8. 2023Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasetscited by 2,581
  9. 2021Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fieldscited by 2,032
  10. 2020pixelNeRF: Neural Radiance Fields from One or Few Imagescited by 1,584
  11. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,567
  12. 2023Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Modelscited by 1,476
38 more
  1. 2017Matterport3D: Learning from RGB-D Data in Indoor Environmentscited by 1,376
  2. 2021Efficient Geometry-aware 3D Generative Adversarial Networkscited by 1,205
  3. 2019Habitat: A Platform for Embodied AI Researchcited by 1,191
  4. 20242D Gaussian Splatting for Geometrically Accurate Radiance Fieldscited by 1,076
  5. 2024Grounding Image Matching in 3D with MASt3Rcited by 983
  6. 2019Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representationscited by 864
  7. 2018MVSNet: Depth Inference for Unstructured Multi-view Stereocited by 788
  8. 2023ScanNet++: A High-Fidelity Dataset of 3D Indoor Scenescited by 764
  9. 2024LGM: Large Multi-View Gaussian Model for High-Resolution 3D Content Creationcited by 739
  10. 2023pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstructioncited by 684
  11. 2023Mip-Splatting: Alias-free 3D Gaussian Splattingcited by 667
  12. 2021Habitat-Matterport 3D Dataset (HM3D): 1000 Large-scale 3D Environments for Embodied AIcited by 592
  13. 2024MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Imagescited by 566
  14. 2023DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Visioncited by 537
  15. 2018Gibson Env: Real-World Perception for Embodied Agentscited by 481
  16. 2019Cascade Cost Volume for High-Resolution Multi-View Stereo and Stereo Matchingcited by 417
  17. 2016Past, Present, and Future of Simultaneous Localization And Mapping: Towards the Robust-Perception Agecited by 411
  18. 2021Mega-NeRF: Scalable Construction of Large-Scale NeRFs for Virtual Fly-Throughscited by 398
  19. 2023Splatter Image: Ultra-Fast Single-View 3D Reconstructioncited by 364
  20. 2024GS-LRM: Large Reconstruction Model for 3D Gaussian Splattingcited by 349
  21. 2023ReconFusion: 3D Reconstruction with Diffusion Priorscited by 318
  22. 2016LIFT: Learned Invariant Feature Transformcited by 306
  23. 2024GRM: Large Gaussian Reconstruction Model for Efficient 3D Reconstruction and Generationcited by 301
  24. 2024gsplat: An Open-Source Library for Gaussian Splattingcited by 285
  25. 20243D Reconstruction with Spatial Memorycited by 277
  26. 2016DSAC - Differentiable RANSAC for Camera Localizationcited by 266
  27. 2020Neural Unsigned Distance Fields for Implicit Function Learningcited by 265
  28. 2023Triplane Meets Gaussian Splatting: Fast and Generalizable Single-View 3D Reconstruction with Transformerscited by 258
  29. 2019Recurrent MVSNet for High-resolution Multi-view Stereo Depth Inferencecited by 254
  30. 2024No Pose, No Problem: Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Imagescited by 248
  31. 2022Panoptic Neural Fields: A Semantic Object-Aware Neural Scene Representationcited by 243
  32. 2022NoPe-NeRF: Optimising Neural Radiance Field with No Pose Priorcited by 235
  33. 2023Visual Geometry Grounded Deep Structure From Motioncited by 227
  34. 2020MetaSDF: Meta-learning Signed Distance Functionscited by 214
  35. 2022Geo-Neus: Geometry-Consistent Neural Implicit Surfaces Learning for Multi-view Reconstructioncited by 212
  36. 2020PatchmatchNet: Learned Multi-View Patchmatch Stereocited by 207
  37. 2021PandaSet: Advanced Sensor Suite Dataset for Autonomous Drivingcited by 190
  38. 2019DIST: Rendering Deep Implicit Signed Distance Function with Differentiable Sphere Tracingcited by 180

50 most cited of 68 All 68 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 71; 2026: 70.

141 papers.

Sort:

By area:

  1. 2025VGGT: Visual Geometry Grounded Transformercited by 1,539
  2. 2025Depth Anything 3: Recovering the Visual Space from Any Viewscited by 692
  3. 2025π3: Permutation-Equivariant Visual Geometry Learningcited by 319
  4. 2025MapAnything: Universal Feed-Forward Metric 3D Reconstructioncited by 314
  5. 2025AnySplat: Feed-forward 3D Gaussian Splatting from Unconstrained Viewscited by 243
  6. 2025ViPE: Video Pose Engine for 3D Geometric Perceptioncited by 170
  7. 2025VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstructioncited by 166
  8. 2025Streaming 4D Visual Geometry Transformercited by 161
  9. 2025Learning from Videos for 3D World: Enhancing MLLMs with 3D Vision Geometry Priorscited by 120
  10. 2025FastVGGT: Training-Free Acceleration of Visual Geometry Transformercited by 98
  11. 2025Easi3R: Estimating Disentangled Motion from DUSt3R Without Trainingcited by 91
  12. 2025OmniWorld: A Multi-Domain and Multi-Modal Dataset for 4D World Modelingcited by 62
38 more
  1. 2025LONG3R: Long Sequence Streaming 3D Reconstructioncited by 45
  2. 2025FantasyWorld: Geometry-Consistent World Modeling via Unified Video and 3D Predictioncited by 38
  3. 2025FlexWorld: Progressively Expanding 3D Scenes for Flexiable-View Synthesiscited by 37
  4. 2025MoVieS: Motion-Aware 4D Dynamic View Synthesis in One Secondcited by 35
  5. 2026V-DPM: 4D Video Reconstruction with Dynamic Point Mapscited by 34
  6. 2026Scal3R: Scalable Test-Time Training for Large-Scale 3D Reconstructioncited by 31
  7. 2025What Is The Best 3D Scene Representation for Robotics? From Geometric to Foundation Modelscited by 29
  8. 2025VGGT4D: Mining Motion Cues in Visual Geometry Transformers for 4D Scene Reconstructioncited by 28
  9. 2025Advances in Feed-Forward 3D Reconstruction and View Synthesis: A Surveycited by 27
  10. 2026VGGT-Omegacited by 26
  11. 2025Block-Sparse Global Attention for Efficient Multi-View Geometry Transformerscited by 23
  12. 2025MoRE: 3D Visual Geometry Reconstruction Meets Mixture-of-Expertscited by 21
  13. 2025PanSt3R: Multi-view Consistent Panoptic Segmentationcited by 20
  14. 2025SAIL-Recon: Large SfM by Augmenting Scene Regression with Localizationcited by 20
  15. 2025AMB3R: Accurate Feed-forward Metric-scale 3D Reconstruction with Backendcited by 20
  16. 2025SteerX: Creating Any Camera-Free 3D and 4D Scenes with Geometric Steeringcited by 15
  17. 2025FreeSplat++: Generalizable 3D Gaussian Splatting for Efficient Indoor Scene Reconstructioncited by 15
  18. 20254D-LRM: Large Space-Time Reconstruction Model From and To Any View at Any Timecited by 14
  19. 2025Revisiting Depth Representations for Feed-Forward 3D Gaussian Splattingcited by 12
  20. 2025Rig3R: Rig-Aware Conditioning for Learned 3D Reconstructioncited by 11
  21. 2025Text-to-3D by Stitching a Multi-view Reconstruction Network to a Video Generatorcited by 11
  22. 2025How Much 3D Do Video Foundation Models Encode?cited by 11
  23. 2025Can Video Diffusion Model Reconstruct 4D Geometry?cited by 10
  24. 2025Review of Feed-forward 3D Reconstruction: From DUSt3R to VGGTcited by 10
  25. 20254D-VGGT: A General Foundation Model with SpatioTemporal Awareness for Dynamic Scene Geometry Estimationcited by 10
  26. 2025Regist3R: Incremental Registration with Stereo Foundation Modelcited by 9
  27. 2025UniGeo: Taming Video Diffusion for Unified Consistent Geometry Estimationcited by 9
  28. 2025ARTDECO: Towards Efficient and High-Fidelity On-the-Fly 3D Reconstruction with Structured Scene Representationcited by 9
  29. 2026UniSH: Unifying Scene and Human Reconstruction in a Feed-Forward Passcited by 9
  30. 2025Unposed 3DGS Reconstruction with Probabilistic Procrustes Mappingcited by 8
  31. 2025PanoSplatt3R: Leveraging Perspective Pretraining for Generalized Unposed Wide-Baseline Panorama Reconstructioncited by 8
  32. 2025PAGE-4D: Disentangled pose and geometry estimation for vggt-4d perceptioncited by 8
  33. 2025One4D: Unified 4D Generation and Reconstruction via Decoupled LoRA Controlcited by 8
  34. 2026Flow4R: Unifying 4D Reconstruction and Tracking with Scene Flowcited by 8
  35. 2026Global Structure-from-Motion Meets Feedforward Reconstructioncited by 8
  36. 2025E3D-Bench: A Benchmark for End-to-End 3D Geometric Foundation Modelscited by 7
  37. 2025Video Perception Models for 3D Scene Synthesiscited by 7
  38. 2025OpenM3D: Open Vocabulary Multi-view Indoor 3D Object Detection without Human Annotationscited by 7

50 most cited of 141 All 141 (JSON)