mlsanityData to 19 Oct 2026

Sparse-input reconstruction

883 papers0 this weekMedium confidence in this grouping: mostly one topic, with blurry edgesMedium confidence in this grouping: mostly one topic, with blurry edges · About

Keywords: reconstruction, view, novel, scene, image

Key facts. Sparse-input reconstruction: 883 papers on arXiv since 2021 (data through 19 Oct 2026). 270 of its 883 mapped papers (31%) cite “DUSt3R: Geometric 3D Vision Made Easy” (2023, arXiv 2312.14132). 42 new papers 1 Oct–19 Oct.

What’s new

1 Oct – 19 Oct 2026

42 new papers, 1 Oct – 19 Oct; 0 this week.

  1. Earlier · 42
  2. ReDT: Resolution-Decoupled Visual Geometry Transformer
  3. LonGaussian: Tracking Foundation Priors for Long-Context Feedforward Gaussian Reconstruction also fits: Semantic 3D scene understanding and robotics
  4. Motion- and Gravity-Aware Feedforward 3D Reconstruction
  5. GeoPano: Towards Geometrically Accurate 3D Reconstruction from a Single Panorama also fits: Neural surface reconstruction
  6. ESC: Escaping 2D Diffusion Ambiguity with Geometric Constraints for Consistent Novel View Synthesis
  7. InfiniSplat v2: Analyzing and Improving Feed-Forward 3DGS from Any Views also fits: 3D Gaussian splatting methods
  8. Remember, Revisit, Revise: Scaffold Memory for Long-Horizon 3D Geometry
  9. MagicGarment: Benchmarking Single-Image Standalone 3D Garment Reconstruction also fits: Avatars
  10. Holi-Spatial 360: Generative Reconstruction from Sparse Panoramas for Spatial Rendering and Intelligence
  11. 3D-VITAL: Visibility-aware Identity Training from Artificial Liftings of 3D Reconstruction model for Cross-View Object Re-Identification
All 42
  1. BAT3R: Robust Online 3D Reconstruction with Bayesian Adaptive State Updates
  2. Graph3R: Unifying View Graph Initialization with Feed-Forward Reconstruction
  3. Sharper Frames, Worse Rooms: Viewpoint Coverage Dominates Image Quality in Feed-Forward 3D Reconstruction
  4. UniForge3D: Joint 3D Reconstruction and Generation in a Shared Anchor-Guided Space also fits: Neural surface reconstruction
  5. Drive3R: Block-wise Streaming Geometric Foundation Models for Autonomous Driving also fits: SLAM and driving scene reconstruction
  6. PURE: Refining Feed-forward 3D Reconstruction with Unlabeled Post-training
  7. Demiurge: Forging Navigable 4D Worlds via Autoregressive Video Rerendering also fits: Dynamic scene reconstruction
  8. NOFACE3D: NOise-scheduled Focused Anchoring and Controlled Enhancement for 3D Generation also fits: Neural surface reconstruction
  9. Zebra3D: Efficient Hybrid Foundation Models For Scalable 3D Reconstruction
  10. LEGO: A Lifting-Free Approach for Exocentric-to-Egocentric Video Generation
  11. Slot3R: Set-Associative Spatial Memory for Streaming 3D Reconstruction
  12. Pose-Free Feed-Forward 3D Inpainting via Learnable Mask Attention and Support Token Refinement
  13. Tetris3D: 3D Scene Generation With Objects That Fit Together
  14. OmniCam: Omni-Camera Trajectory Generation via Geometry-Grounded Pose Token Learning also fits: Semantic 3D scene understanding and robotics
  15. Hardware-aware Calibrated Clustered Attention for Efficient Visual Geometric Transformers
  16. RDGSplat: Render-Dedicated Geometry for Novel View Synthesis
  17. Revar3r: gauge-aware perturbation uncertainty for feed-forward 3d reconstruction
  18. AIMS: Anchor-Integrated Multi-View Synthesis for Scalable Novel View Rendering
  19. Less Context, Better Geometry: Masked Geometric Encoder for Robust 3D Foundation Models
  20. VGGT-Bridge: Beyond Sequential Pose Graphs via Coarse-Stride Skip Edges
  21. Every View Counts: View-Consistent Panoptic Quality for Multi-view Panoptic Segmentation
  22. NAMVIS: Next-Scale Autoregressive Multi-View Image Synthesis
  23. Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis
  24. Feedforward Novel View Synthesis for Heterogeneous Cameras
  25. I2CD: Direct Image-to-Convex Decomposition for Simulation-Ready Collision Geometry also fits: Semantic 3D scene understanding and robotics
  26. T3lescope: Arbitrary-Resolution High-Fidelity Generative Surface Reconstruction from Images
  27. CLoSeR: Closing the Loop for Long-Context Streaming Reconstruction
  28. DecomVoxel: Harnessing 3D-Native Priors with Guided In-situ Denoising Optimization for Decompositional Scene Reconstruction
  29. ARROW: Arbitrary Reconstruction and Tracking of 4D Observations in the Wild
  30. MVDG: Efficient Multi-view 3D Disambiguation on Unconstrained Real-World Images also fits: Multi-view stereo and depth estimation
  31. HierGF: Hierarchical Gaussian Fields via Geometry-perception Message Passing for Sparse-view 3D Reconstruction also fits: 3D Gaussian splatting methods
  32. VASC: Value-Aware Sparse Attention with Cross-Layer Memory for Efficient 3D Reconstruction

Foundations

Most cited by papers here

  1. DUSt3R: Geometric 3D Vision Made Easy2023cited by 270 here
  2. Zero-1-to-3: Zero-shot One Image to 3D Object2023cited by 235 here
  3. Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction2021cited by 164 here
  4. LRM: Large Reconstruction Model for Single Image to 3D2023cited by 138 here
  5. Continuous 3D Perception Model with Persistent State2025cited by 127 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2021Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fieldscited by 127 here
  2. 2021Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fieldscited by 144 here
  3. 2022DreamFusion: Text-to-3D using 2D Diffusioncited by 207 here
  4. 2022Magic3D: High-Resolution Text-to-3D Content Creationcited by 101 here
  5. 2023Zero-1-to-3: Zero-shot One Image to 3D Objectcited by 235 here
  6. 2023One-2-3-45: Any Single Image to 3D Mesh in 45 Seconds without Per-Shape Optimizationcited by 115 here
  7. 2023LRM: Large Reconstruction Model for Single Image to 3Dcited by 138 here
  8. 2024CAT3D: Create Anything in 3D with Multi-View Diffusion Modelscited by 105 here
  9. 2025Continuous 3D Perception Model with Persistent Statecited by 127 here
15 more it builds on, not in this chain
  1. 2021IBRNet: Learning Multi-View Image-Based Renderingcited by 122 here
  2. 2021MVSNeRF: Fast Generalizable Radiance Field Reconstruction from Multi-View Stereocited by 117 here
  3. 2021NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstructioncited by 120 here
  4. 2021Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstructioncited by 164 here
  5. 2023ProlificDreamer: High-Fidelity and Diverse Text-to-3D Generation with Variational Score Distillationcited by 103 here
  6. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 383 here
  7. 2023ScanNet++: A High-Fidelity Dataset of 3D Indoor Scenescited by 100 here
  8. 2023SyncDreamer: Generating Multiview-consistent Images from a Single-view Imagecited by 126 here
  9. 2023Wonder3D: Single Image to 3D using Cross-Domain Diffusioncited by 105 here
  10. 2023ReconFusion: 3D Reconstruction with Diffusion Priorscited by 86 here
  11. 2023pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstructioncited by 106 here
  12. 2023DUSt3R: Geometric 3D Vision Made Easycited by 270 here
  13. 2023DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Visioncited by 115 here
  14. 2024MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Imagescited by 94 here
  15. 2025Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Passcited by 106 here

The 24 most cited of 1,859.

Browse

30 of 883, ranked 6–35

  1. 6SyncDreamer: Generating Multiview-consistent Images from a Single-view Image2023126
  2. 7IBRNet: Learning Multi-View Image-Based Rendering2021122
  3. 8DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Vision2023115
  4. 9Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass2025106
  5. 10CAT3D: Create Anything in 3D with Multi-View Diffusion Models2024105
  6. 11Wonder3D: Single Image to 3D using Cross-Domain Diffusion2023105
  7. 12ScanNet++: A High-Fidelity Dataset of 3D Indoor Scenes2023100
  8. 13ViewCrafter: Taming Video Diffusion Models for High-fidelity Novel View Synthesis202486
  9. 143D Reconstruction with Spatial Memory202484
  10. 15Generative Novel View Synthesis with 3D-Aware Diffusion Models202383
All 30
  1. 16Novel View Synthesis with Diffusion Models202283
  2. 17Scene Representation Transformer: Geometry-Free Novel View Synthesis Through Set-Latent Scene Representations202174
  3. 18SV3D: Novel Multi-view Synthesis and 3D Generation from a Single Image using Latent Video Diffusion202467
  4. 19Magic123: One Image to High-Quality 3D Object Generation Using Both 2D and 3D Diffusion Priors202367
  5. 20LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias202465
  6. 21MapAnything: Universal Feed-Forward Metric 3D Reconstruction202561
  7. 22RealFusion: 360deg Reconstruction of Any Object from a Single Image202358
  8. 23One-2-3-45++: Fast Single Image to 3D Objects with Consistent Multi-View Generation and 3D Diffusion202355
  9. 24FLARE: Feed-forward Geometry, Appearance and Camera Estimation from Uncalibrated Sparse Views202553
  10. 25OmniObject3D: Large-Vocabulary 3D Object Dataset for Realistic Perception, Reconstruction and Generation202352
  11. 26TTT3R: 3D Reconstruction as Test-Time Training202551
  12. 27Make-It-3D: High-Fidelity 3D Creation from A Single Image with Diffusion Prior202350
  13. 28MUSt3R: Multi-view Network for Stereo 3D Reconstruction202546
  14. 29SparseFusion: Distilling View-conditioned Diffusion for 3D Reconstruction202246
  15. 30Streaming 4D Visual Geometry Transformer202545
  16. 31DMV3D: Denoising Multi-View Diffusion using 3D Large Reconstruction Model202343
  17. 32RGBD Objects in the Wild: Scaling Real-World 3D Object Learning from RGB-D Videos202441
  18. 33MV-DUSt3R+: Single-Stage Scene Reconstruction from Sparse Views In 2 Seconds202439
  19. 34PF-LRM: Pose-Free Large Reconstruction Model for Joint Pose and Shape Prediction202339
  20. 35ReconX: Reconstruct Any Scene from Sparse Views with Video Diffusion Model202437