Awesome Feed-Forward 3D

An curated list for feed-forward 3D scene modeling, including research directions, datasets, and applications.

Table of Contents

• Research Directions
  • Feature Enhancement
    • Advanced Encoding Architectures
    • Cross-View Fusion
    • Integration of Visual Foundation Models
  • Geometry-aware Improvement
    • Explicit Geometric Aggregation
    • Refining Predicted 3D Scenes
    • Pose-Free Reconstruction
    • Pre-trained Geometric Guidance
  • Model Efficiency
    • Feature Efficiency
    • Representation Compaction
  • Data & Visual Augmentation
    • Data Augmentation
    • Visual Augmentation
  • Temporal-aware Models
    • Online Streaming
    • Offline Processing
    • Interactive Modeling
    • Specialized Tasks
• Datasets and Benchmarks
  • Geometry Oriented
  • Visual Oriented
  • Mixed
• Applications
  • Autonomous Driving
  • Robotics
  • SfM & SLAM
  • Scene Understanding
  • Video Generation
  • Others

Taxonomy

Research Directions

Feature Enhancement

Advanced Encoding Architectures

• pixelNeRF: Neural Radiance Fields from One or Few Images. [📄 Paper | 💻 Code]
• IBRNet: Learning Multi-View Image-Based Rendering. [📄 Paper | 💻 Code｜🌐 Project Page]
• Splatter Image: Ultra-Fast Single-View 3D Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• Convolutional Occupancy Networks. [📄 Paper | 💻 Code｜🌐 Project Page]
• Light Field Networks: Neural Scene Representations with Single-Evaluation Rendering. [📄 Paper | 💻 Code｜🌐 Project Page]
• Learned Initializations for Optimizing Coordinate-Based Neural Representations. [📄 Paper | 💻 Code｜🌐 Project Page]
• Neural Rays for Occlusion-aware Image-based Rendering. [📄 Paper | 💻 Code｜🌐 Project Page]
• ${C}^{3}$-GS: Learning Context-aware, Cross-dimension, Cross-scale Feature for Generalizable Gaussian Splatting. [📄 Paper | 💻 Code]
• Scene Representation Transformer: Geometry-Free Novel View Synthesis Through Set-Latent Scene Representations. [📄 Paper | 💻 Code｜🌐 Project Page]
• VisionNeRF: Vision Transformer for NeRF-Based View Synthesis from a Single Input Image. [📄 Paper | 💻 Code｜🌐 Project Page]
• RePAST: Relative Pose Attention Scene Representation Transformer. [📄 Paper]
• Is Attention All That NeRF Needs? [📄 Paper | 💻 Code｜🌐 Project Page]
• Large Reconstruction Model (LRM).
• Instant3D: Fast Text-to-3D with Sparse-view Generation and Large Reconstruction Model.
• TripoSR: Fast 3D Object Reconstruction from a Single Image. [📄 Paper]
• GRM: Large Gaussian Reconstruction Model for Efficient 3D Reconstruction and Generation.
• GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting. [📄 Paper | 💻 Code｜🌐 Project Page]
• MeshLRM: Large Reconstruction Model for High-Quality Meshes.
• MeshFormer: High-Quality Mesh Generation with 3D-Guided Reconstruction Model.
• Flex3D: Feed-Forward 3D Generation With Flexible Reconstruction Model And Input View Curation.
• LVSM: A Fully Data-Driven Approach to Novel View Synthesis.
• Depth Anything 3: Recovering the Visual Space from Any Views.
• Gamba: Marry Gaussian Splatting With Mamba for Single-View 3D Reconstruction.
• MVGamba: Unify 3D Content Generation as State Space Sequence Modeling.
• Long-LRM: Long-sequence Large Reconstruction Model for Wide-coverage Gaussian Splats. [📄 Paper | 💻 Code｜🌐 Project Page]

Cross-View Fusion

• AttnRend: Learning to Render Novel Views from Wide-Baseline Stereo Pairs. [📄 Paper | 💻 Code｜🌐 Project Page]
• Epipolar-Free 3D Gaussian Splatting for Generalizable Novel View Synthesis. [📄 Paper | 💻 Code｜🌐 Project Page]
• LGM: Large Multi-View Gaussian Model for High-Resolution 3D Content Creation.
• DUSt3R: Geometric 3D Vision Made Easy. [📄 Paper | 💻 Code｜🌐 Project Page]
• Grounding Image Matching in 3D with MASt3R. [📄 Paper | 💻 Code｜🌐 Project Page]
• MV-DUSt3R: Multi-View Dense Stereo 3D Reconstruction.
• MV-DUSt3R+: Single-Stage Scene Reconstruction from Sparse Views In 2 Seconds. [📄 Paper | 💻 Code｜🌐 Project Page]
• PreF3R: Pose-Free Feed-Forward 3D Gaussian Splatting from Variable-length Image Sequence. [📄 Paper | 💻 Code｜🌐 Project Page]
• 3D Reconstruction with Spatial Memory. [📄 Paper | 💻 Code｜🌐 Project Page]
• Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass. [📄 Paper | 💻 Code｜🌐 Project Page]
• MUSt3R: Multi-view Network for Stereo 3D Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• WinT3R: Window-Based Streaming Reconstruction with Camera Token Pool. [📄 Paper]
• Continuous 3D Perception Model with Persistent State. [📄 Paper | 💻 Code｜🌐 Project Page]
• VolSplat: Rethinking Feed-Forward 3D Gaussian Splatting with Voxel-Aligned Prediction. [📄 Paper]
• G-CUT3R: Guided 3D Reconstruction with Camera and Depth Prior Integration. [📄 Paper]
• TTT3R: 3D Reconstruction as Test-Time Training. [📄 Paper]
• Point3R: Streaming 3D Reconstruction with Explicit Spatial Pointer Memory. [📄 Paper]
• VGGT: Visual Geometry Grounded Transformer. [📄 Paper | 💻 Code｜🌐 Project Page]
• iLRM: An Iterative Large 3D Reconstruction Model. [📄 Paper | 💻 Code｜🌐 Project Page]
• Dens3r: A Foundation Model for 3D Geometry Prediction. [📄 Paper]
• MoRE: 3D Visual Geometry Reconstruction Meets Mixture-of-Experts. [📄 Paper]
• Uni3R: Unified 3D Reconstruction and Semantic Understanding via Generalizable Gaussian Splatting from Unposed Multi-view Images. [📄 Paper]
• Flow3r: Factored Flow Prediction for Scalable Visual Geometry Learning. [📄 Paper]
• Gen3R: 3D Scene Generation Meets Feed-Forward Reconstruction. [📄 Paper]
• NOVA3R: Non-pixel-aligned Visual Transformer for Amodal 3D Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• IncVGGT: Incremental VGGT for Memory-Bounded Long-Range 3D Reconstruction.
• ZipMap: Linear-Time Stateful 3D Reconstruction with Test-Time Training. [📄 Paper]
• LoGeR: Long-Context Geometric Reconstruction with Hybrid Memory. [📄 Paper]
• tttLRM: Test-Time Training for Long Context and Autoregressive 3D Reconstruction. [📄 Paper]
• VGG-T3: Offline Feed-Forward 3D Reconstruction at Scale. [📄 Paper]

Integration of Visual Foundation Models

• DUSt3R: Geometric 3D Vision Made Easy. [📄 Paper | 💻 Code｜🌐 Project Page]
• Mono3R: Exploiting Monocular Cues for Geometric 3D Reconstruction. [📄 Paper]
• Feat2GS: Probing Visual Foundation Models with Gaussian Splatting. [📄 Paper | 💻 Code｜🌐 Project Page]
• CATSplat: Context-Aware Transformer with Spatial Guidance for Generalizable 3D Gaussian Splatting from A Single-View Image. [📄 Paper]

Geometry-aware Improvement

Explicit Geometric Aggregation

• MVSNeRF: Fast Generalizable Radiance Field Reconstruction from Multi-View Stereo. [📄 Paper | 💻 Code｜🌐 Project Page]
• Stereo Radiance Fields (SRF): Learning View Synthesis for Sparse Views of Novel Scenes. [📄 Paper | 💻 Code｜🌐 Project Page]
• GeoNeRF: Generalizing NeRF with Geometry Priors. [📄 Paper | 💻 Code｜🌐 Project Page]
• BoostMVSNeRFs: Boosting MVS-based NeRFs to Generalizable View Synthesis in Large-Scale Scenes.
• Generalizable Patch-Based Neural Rendering. [📄 Paper | 💻 Code｜🌐 Project Page]
• MatchNeRF: Explicit Correspondence Matching for Generalizable Neural Radiance Fields. [📄 Paper | 💻 Code｜🌐 Project Page]
• GTA: A Geometry-Aware Attention Mechanism for Multi-View Transformers. [📄 Paper | 💻 Code｜🌐 Project Page]
• MuRF: Multi-Baseline Radiance Fields. [📄 Paper | 💻 Code｜🌐 Project Page]
• SparseNeuS: Fast Generalizable Neural Surface Reconstruction from Sparse Views.
• VolRecon: Volume Rendering of Signed Ray Distance Functions for Generalizable Multi-View Reconstruction.
• ReTR: Modeling Depth Distribution for Generalizable Surface Reconstruction.
• UFORecon: Generalizable Sparse-View Surface Reconstruction from Arbitrary and Unfavorable Sets.
• SurfaceSplat: Connecting Surface Reconstruction and Gaussian Splatting. [📄 Paper]
• RenderFormer: Transformer-based Neural Rendering of Triangle Meshes with Global Illumination. [📄 Paper | 💻 Code｜🌐 Project Page]
• AGG: Amortized Generative 3D Gaussians for Single Image to 3D. [📄 Paper]
• TGS: Triplane Meets Gaussian Splatting: Fast and Generalizable Single-View 3D Reconstruction with Transformers.
• LaRa: Efficient Large-Baseline Radiance Fields. [📄 Paper | 💻 Code｜🌐 Project Page]
• MeshSplat: Generalizable Sparse-View Surface Reconstruction via Gaussian Splatting. [📄 Paper]
• TranSplat: Geometry-Aware Feed-Forward Gaussian Splatting with Transformation Consistency.
• H3R: Hybrid Multi-view Correspondence for Generalizable 3D Reconstruction. [📄 Paper]
• MuGS: Multi-Baseline Generalizable Gaussian Splatting Reconstruction.
• pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images. [📄 Paper | 💻 Code｜🌐 Project Page]
• MVSGaussian: Fast Generalizable Gaussian Splatting Reconstruction from Multi-View Stereo. [📄 Paper | 💻 Code｜🌐 Project Page]

Refining Predicted 3D Scenes

• FreeSplat: Generalizable 3D Gaussian Splatting Towards Free-View Synthesis of Indoor Scenes. [📄 Paper | 💻 Code｜🌐 Project Page]
• HiSplat: Hierarchical 3D Gaussian Splatting for Generalizable Sparse-View Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• PixelGaussian: Generalizable 3D Gaussian Reconstruction from Arbitrary Views. [📄 Paper | 💻 Code｜🌐 Project Page]
• Gaussian Graph Network: Learning Efficient and Generalizable Gaussian Representations from Multi-view Images. [📄 Paper | 💻 Code｜🌐 Project Page]
• Generative Densification: Learning to Densify Gaussians for High-Fidelity Generalizable 3D Reconstruction.
• G3R: Gradient Guided Generalizable Reconstruction.

Pose-Free Reconstruction

• LEAP: Liberate Sparse-view 3D Modeling from Camera Poses. [📄 Paper | 💻 Code｜🌐 Project Page]
• Splatt3R: Zero-shot Gaussian Splatting from Uncalibrated Image Pairs. [📄 Paper | 💻 Code｜🌐 Project Page]
• No Pose, No Problem: Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Images. [📄 Paper | 💻 Code｜🌐 Project Page]
• PF3plat: Pose-Free Feed-Forward 3D Gaussian Splatting. [📄 Paper | 💻 Code｜🌐 Project Page]
• FreeSplatter: Pose-free Gaussian Splatting for Sparse-view 3D Reconstruction. [📄 Paper | 💻 Code]
• FLARE: Feed-forward Geometry, Appearance and Camera Estimation from Uncalibrated Sparse Views. [📄 Paper | 💻 Code｜🌐 Project Page]
• Pow3R: Empowering Unconstrained 3D Reconstruction with Camera and Scene Priors. [📄 Paper | 💻 Code｜🌐 Project Page]
• RegGS: Unposed Sparse Views Gaussian Splatting with 3DGS Registration. [📄 Paper | 💻 Code｜🌐 Project Page]
• UFV-Splatter: Pose-Free Feed-Forward 3D Gaussian Splatting Adapted to Unfavorable Views. [📄 Paper | 💻 Code｜🌐 Project Page]
• π^3: Scalable Permutation-Equivariant Visual Geometry Learning. [📄 Paper | 💻 Code｜🌐 Project Page]
• AnySplat: Feed-forward 3D Gaussian Splatting from Unconstrained Views. [📄 Paper]
• No Pose at All: Self-Supervised Pose-Free 3D Gaussian Splatting from Sparse Views.
• SPFSplatV2: Efficient Self-Supervised Pose-Free 3D Gaussian Splatting from Sparse Views. [📄 Paper]
• PLANA3R: Zero-shot Metric Planar 3D Reconstruction via Feed-forward Planar Splatting.
• YoNoSplat: You Only Need One Model for Feedforward 3D Gaussian Splatting. [📄 Paper]

Pre-trained Geometric Guidance

• DepthSplat: Connecting Gaussian Splatting and Depth. [📄 Paper | 💻 Code｜🌐 Project Page]
• Flash3D: Feed-Forward Generalisable 3D Scene Reconstruction from a Single Image. [📄 Paper | 💻 Code｜🌐 Project Page]
• Niagara: Normal-Integrated Geometric Affine Field for Scene Reconstruction from a Single View. [📄 Paper | 💻 Code｜🌐 Project Page]
• Revisiting Depth Representations for Feed-Forward 3D Gaussian Splatting. [📄 Paper | 💻 Code｜🌐 Project Page]
• Fin3R: Fine-tuning Feed-forward 3D Reconstruction Models via Monocular Knowledge Distillation. [📄 Paper]
• JointSplat: Joint Depth and Flow Priors for Feed-Forward 3D Gaussian Splatting. [📄 Paper]

Model Efficiency

Feature Efficiency

• Efficient Neural Radiance Fields for Interactive Free-viewpoint Video. [📄 Paper | 💻 Code｜🌐 Project Page]
• ProNeRF: Learning Efficient Projection-Aware Ray Sampling for Fine-Grained Implicit Neural Radiance Fields. [📄 Paper | 💻 Code｜🌐 Project Page]
• TinySplat: Feedforward Approach for Generating Compact 3D Scene Representation. [📄 Paper]
• ZPressor: Bottleneck-Aware Compression for Scalable Feed-Forward 3DGS. [📄 Paper | 💻 Code｜🌐 Project Page]
• FastVGGT: Training-Free Acceleration of Visual Geometry Transformer. [📄 Paper | 💻 Code｜🌐 Project Page]
• Quantized Visual Geometry Grounded Transformer. [📄 Paper]
• Faster VGGT with Block-Sparse Global Attention. [📄 Paper]
• Evict3R: Training-Free Token Eviction for Memory-Bounded Streaming Visual Geometry Transformers. [📄 Paper]
• LiteVGGT: Boosting Vanilla VGGT via Geometry-aware Cached Token Merging. [📄 Paper]
• Speed3R: Sparse Feed-forward 3D Reconstruction Models. [📄 Paper]
• SR3R: Rethinking Super-Resolution 3D Reconstruction With Feed-Forward Gaussian Splatting. [📄 Paper]

Representation Compaction

• Gaussian Graph Network: Learning Efficient and Generalizable Gaussian Representations from Multi-view Images. [📄 Paper | 💻 Code｜🌐 Project Page]
• PixelGaussian: Generalizable 3D Gaussian Reconstruction from Arbitrary Views. [📄 Paper | 💻 Code｜🌐 Project Page]
• FreeSplat++: Generalizable 3D Gaussian Splatting for Efficient Indoor Scene Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• LongSplat: Online Generalizable 3D Gaussian Splatting from Long Sequence Images. [📄 Paper | 💻 Code｜🌐 Project Page]

Data & Visual Augmentation

Data Augmentation

• MegaSynth: Scaling Up 3D Scene Reconstruction with Synthesized Data. [📄 Paper | 💻 Code｜🌐 Project Page]
• Puzzles: Unbounded Video-Depth Augmentation for Scalable End-to-End 3D Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• Aug3D: Augmenting Large Scale Outdoor Datasets for Generalizable Novel View Synthesis. [📄 Paper]
• MVBoost: Boost 3D Reconstruction with Multi-View Refinement.

Visual Augmentation

• MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views. [📄 Paper | 💻 Code｜🌐 Project Page]
• latentSplat: Autoencoding Variational Gaussians for Fast Generalizable 3D Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• ProSplat: Improved Feed-Forward 3D Gaussian Splatting for Wide-Baseline Sparse Views. [📄 Paper]
• Difix3D+: Improving 3D Reconstructions with Single-Step Diffusion Models.
• Reconstruct, Inpaint, Finetune: Dynamic Novel-view Synthesis from Monocular Videos. [📄 Paper]

Temporal-aware Models

Online Streaming

• StreamSplat: Towards Online Dynamic 3D Reconstruction from Uncalibrated Video Streams. [📄 Paper | 💻 Code]
• Continuous 3D Perception Model with Persistent State. [📄 Paper | 💻 Code｜🌐 Project Page]
• DGS-LRM: Real-Time Deformable 3D Gaussian Reconstruction From Monocular Videos. [📄 Paper ｜🌐 Project Page]
• Stream3R: Scalable Sequential 3D Reconstruction with Causal Transformer. [📄 Paper]
• LongStream: Long-Sequence Streaming Autoregressive Visual Geometry. [📄 Paper]

Offline Processing

• L4GM: Large 4D Gaussian Reconstruction Model.
• 4D-LRM: Large Space-Time Reconstruction Model From and To Any View at Any Time. [📄 Paper]
• MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion. [📄 Paper | 💻 Code｜🌐 Project Page]
• Easi3R: Estimating Disentangled Motion from DUSt3R Without Training. [📄 Paper | 💻 Code｜🌐 Project Page]
• 4DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videos. [📄 Paper | 💻 Code｜🌐 Project Page]
• MoVieS: Motion-Aware 4D Dynamic View Synthesis in One Second. [📄 Paper | 💻 Code｜🌐 Project Page]
• 4Real-Video-V2: Fused View-Time Attention and Feedforward Reconstruction for 4D Scene Generation. [📄 Paper ｜🌐 Project Page]
• MonoFusion: Sparse-View 4D Reconstruction via Monocular Fusion. [📄 Paper | 💻 Code｜🌐 Project Page]
• Self-Supervised Monocular 4D Scene Reconstruction for Egocentric Videos. [📄 Paper]
• Feed-forward Bullet-Time Reconstruction of Dynamic Scenes from Monocular Videos. [📄 Paper]

Interactive Modeling

• PIXIE: Physics from Pixels for Interactive Feed-Forward Scene Modeling. [📄 Paper]
• PhysGM: Physical Gaussian Modeling for Interactive 3D Scene Editing. [📄 Paper]

Specialized Tasks

• DAS3R: Dynamics-Aware Gaussian Splatting for Static Scene Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• St4RTrack: Simultaneous 4D Reconstruction and Tracking in the World.

Datasets and Benchmarks

Geometry Oriented

• DTU: Large Scale Multi-view Stereopsis Evaluation. [📄 Paper | 🌐 Project Page]
• 7Scenes: Scene Coordinate Regression Forests for Camera Relocalization in RGB-D Images. [📄 Paper | 🌐 Project Page]

Visual Oriented

• NeRF-Synthetic: NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• Neural 3D Mesh Renderer (NMR): Neural 3D Mesh Renderer. [📄 Paper | 💻 Code]
• CelebA: Deep Learning Face Attributes in the Wild. [📄 Paper | 🌐 Project Page | 🔗Data Link]
• Consistent4D: Consistent 360° Dynamic Object Generation from Monocular Video. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• NYUv2: Indoor Segmentation and Support Inference from RGBD Images. [📄 Paper | 🌐 Project Page | 🔗Data Link]
• Habitat: A Platform for Embodied AI Research. [📄 Paper | 💻 Code | 🔗Data Link]
• Hot3D: Hand and Object Tracking in 3D from Egocentric Multi-view Videos. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• ACID: Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Image. [📄 Paper ｜🌐 Project Page | 🔗Data Link]
• ENeRF-Outdoor: Efficient Neural Radiance Fields for Interactive Free-viewpoint Video. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• LLFF: Local Light Field Fusion: Practical View Synthesis with Prescriptive Sampling Guidelines. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• Neural3DV: Neural 3D Video Synthesis from Multi-view Video. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• EgoExo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectives. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• DAVIS: A Benchmark Dataset and Evaluation Methodology for Video Object Segmentation. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• Youtube-VOS: Sequence-to-sequence Video Object Segmentation. [📄 Paper ｜🌐 Project Page | 🔗Data Link]
• DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Vision. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• RealEstate10K: Stereo Magnification: Learning View Synthesis Using Multiplane Images. [📄 Paper ｜🌐 Project Page | 🔗Data Link]

Mixed

• Google Scanned Objects (GSO): A High-Quality Dataset of 3D Scanned Household Items. [📄 Paper | 🌐 Project Page | 🔗Data Link]
• OmniObject3D: Large-Vocabulary 3D Object Dataset for Realistic Perception, Reconstruction and Generation. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• CO3D: Common Objects in 3D: Large-Scale Learning and Evaluation of Real-Life 3D Category Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• WildRGBD: RGBD Objects in the Wild: Scaling Real-World 3D Object Learning from RGB-D Videos. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• ShapeNet: An Information-Rich 3D Model Repository. [📄 Paper ｜🌐 Project Page | 🔗Data Link]
• MVImgNet: A Large-Scale Dataset of Multi-View Images. [📄 Paper | 💻 Code | 🔗Data Link]
• Objaverse: A Universe of Annotated 3D Objects. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• Objaverse-XL: A Universe of 10M+ 3D Objects. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• DeepVoxels: Learning Persistent 3D Feature Embeddings. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• MultiShapeNet: Scene Representation Transformer: Geometry-Free Novel View Synthesis Through Set-Latent Scene Representations. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• Amazon Berkeley Objects (ABO): Dataset and Benchmarks for Real-World 3D Object Understanding. [📄 Paper ｜🌐 Project Page | 🔗Data Link]
• Replica: A Digital Replica of Indoor Spaces. [📄 Paper | 💻 Code | 🔗Data Link]
• TUM RGBD: Evaluating Egomotion and Structure-from-Motion Approaches Using the TUM RGB-D Benchmark. [📄 Paper ｜🌐 Project Page | 🔗Data Link]
• Matterport3D: Learning from RGB-D Data in Indoor Environments. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• Hypersim: A Photorealistic Synthetic Dataset for Holistic Indoor Scene Understanding. [📄 Paper | 💻 Code | 🔗Data Link]
• ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• ScanNet++: A High-Fidelity Dataset of 3D Indoor Scenes. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• ARKitScenes: A Diverse Real-World Dataset for 3D Indoor Scene Understanding Using Mobile RGB-D Data. [📄 Paper | 💻 Code | 🔗Data Link]
• Virtual KITTI 2. [📄 Paper ｜🌐 Project Page | 🔗Data Link]
• Spring: A High-Resolution High-Detail Dataset and Benchmark for Scene Flow, Optical Flow and Stereo. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• MegaDepth: Learning Single-View Depth Prediction from Internet Photos. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• PointOdyssey: A Large-Scale Synthetic Dataset for Long-Term Point Tracking. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• TartanAir: A Dataset to Push the Limits of Visual SLAM. [📄 Paper ｜🌐 Project Page | 🔗Data Link]
• Waymo: Scalability in Perception for Autonomous Driving. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• nuScenes: A Multimodal Dataset for Autonomous Driving. [📄 Paper | 🌐 Project Page | 🔗Data Link]
• Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• Tanks and Temples: Benchmarking Large-Scale Scene Reconstruction. [📄 Paper | 🌐 Project Page]
• ETH3D: A Multi-View Stereo Benchmark with High-Resolution Images and Multi-Camera Videos. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]
• BlendedMVS: A Large-Scale Dataset for Generalized Multi-View Stereo Networks. [📄 Paper | 💻 Code | 🔗Data Link]
• DyCheck: Dynamic Gaussian Marbles for Novel View Synthesis of Casual Monocular Videos. [📄 Paper | 💻 Code｜🌐 Project Page | 🔗Data Link]

Applications

Autonomous Driving

• Driv3R: Learning Dense 4D Reconstruction for Autonomous Driving. [📄 Paper | 💻 Code｜🌐 Project Page]
• InfiniCube: Unbounded and Controllable Dynamic 3D Driving Scene Generation with World-Guided Video Models. [📄 Paper | 💻 Code｜🌐 Project Page]
• DrivingRecon: Large 4D Gaussian Reconstruction Model for Autonomous Driving. [📄 Paper | 💻 Code]
• GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control. [📄 Paper | 💻 Code｜🌐 Project Page]
• DrivingForward: Feed-forward 3D Gaussian Splatting for Driving Scene Reconstruction from Flexible Surround-view Input. [📄 Paper | 💻 Code]
• STORM: Spatio-Temporal Reconstruction Model for Large-Scale Outdoor Scenes. [📄 Paper | 💻 Code｜🌐 Project Page]
• SCube: Instant Large-Scale Scene Reconstruction using VoxSplats. [📄 Paper | 💻 Code｜🌐 Project Page]
• EVolSplat: Efficient Volume-based Gaussian Splatting for Urban View Synthesis. [📄 Paper | 💻 Code｜🌐 Project Page]
• Omni-Scene: Omni-Gaussian Representation for Ego-Centric Sparse-View Scene Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]
• BEV-GS: Feed-forward Gaussian Splatting in Bird's-Eye-View for Road Reconstruction. [📄 Paper | 💻 Code]
• Efficient Depth-guided Urban View Synthesis. [📄 Paper | 💻 Code｜🌐 Project Page]
• DriveGen3D: Boosting Feed-Forward Driving Scene Generation with Efficient Video Diffusion. [📄 Paper｜🌐 Project Page]
• WorldSplat: Gaussian-Centric Feed-Forward 4D Scene Generation for Autonomous Driving. [📄 Paper | 💻 Code｜🌐 Project Page]

Robotics

Manipulation

• GraspNeRF: Multiview-based 6-DoF Grasp Detection for Transparent and Specular Objects Using Generalizable NeRF. [📄 Paper | 💻 Code｜🌐 Project Page]
• ManiGaussian: Dynamic Gaussian Splatting for Multi-task Robotic Manipulation. [📄 Paper | 💻 Code｜🌐 Project Page]
• ManiGaussian++: General Robotic Bimanual Manipulation with Hierarchical Gaussian World Model. [📄 Paper | 💻 Code]
• Query-based Semantic Gaussian Field for Scene Representation in Reinforcement Learning. [📄 Paper]
• GAF: Gaussian Action Field as a Dynamic World Model for Robotic Manipulation. [📄 Paper｜🌐 Project Page]
• GaussianGrasper: 3D Language Gaussian Splatting for Open-Vocabulary Robotic Grasping. [📄 Paper | 💻 Code]
• EmbodiedSplat: Personalized Real-to-Sim-to-Real Navigation with Gaussian Splats from a Mobile Device. [📄 Paper｜🌐 Project Page]

Navigation

• IGL-Nav: Incremental 3D Gaussian Localization for Image-goal Navigation. [📄 Paper | 💻 Code｜🌐 Project Page]
• VR-Robo: A Real-to-Sim-to-Real Framework for Visual Robot Navigation and Locomotion. [📄 Paper | 💻 Code｜🌐 Project Page]
• GS-LTS: 3D Gaussian Splatting-Based Adaptive Modeling for Long-Term Service Robots. [📄 Paper｜🌐 Project Page]
• UnitedVLN: Generalizable Gaussian Splatting for Continuous Vision-Language Navigation. [📄 Paper]

SfM & SLAM

SFM

• Visual Geometry Grounded Deep Structure From Motion. [📄 Paper | 💻 Code｜🌐 Project Page]
• Light3R-SfM: Towards Feed-forward Structure-from-Motion. [📄 Paper｜🌐 Project Page]
• Mast3r-sfm: A Fully-Integrated Solution for Unconstrained Structure-from-Motion. [📄 Paper | 💻 Code]
• Regist3R: Incremental Registration with Stereo Foundation Model. [📄 Paper | 💻 Code]
• VGGT-Long: Chunk it, Loop it, Align it -- Pushing VGGT's Limits on Kilometer-scale Long RGB Sequences. [📄 Paper | 💻 Code]
• Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass. [📄 Paper | 💻 Code｜🌐 Project Page]
• FLARE: Feed-forward Geometry, Appearance and Camera Estimation from Uncalibrated Sparse Views. [📄 Paper | 💻 Code｜🌐 Project Page]

SLAM

• MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors. [📄 Paper | 💻 Code｜🌐 Project Page]
• SLAM3R. [📄 Paper | 💻 Code]
• VGGT-SLAM. [📄 Paper | 💻 Code]
• ARTDECO: Towards Efficient and High-Fidelity On-the-Fly 3D Reconstruction with Structured Scene Representation. [📄 Paper]
• EC3R-SLAM: Efficient and Consistent Monocular Dense SLAM with Feed-Forward 3D Reconstruction. [📄 Paper]
• MASt3R-Fusion: Integrating Feed-Forward Visual Model with IMU, GNSS for High-Functionality SLAM. [📄 Paper]
• ViSTA-SLAM: Visual SLAM with Symmetric Two-view Association. [📄 Paper | 💻 Code｜🌐 Project Page]

Scene Understanding

Semantic

• SLGaussian: Fast Language Gaussian Splatting in Sparse Views. [📄 Paper]
• GSemSplat: Generalizable Semantic 3D Gaussian Splatting from Uncalibrated Image Pairs. [📄 Paper]
• SegMASt3R: Geometry Grounded Segment Matching. [📄 Paper｜🌐 Project Page]
• PartField: Learning 3D Feature Fields for Part Segmentation and Beyond. [📄 Paper | 💻 Code｜🌐 Project Page]
• Semantic Gaussians: Open-Vocabulary Scene Understanding with 3D Gaussian Splatting. [📄 Paper | 💻 Code｜🌐 Project Page]
• SemanticSplat: Feed-Forward 3D Scene Understanding with Language-Aware Gaussian Fields. [📄 Paper | 💻 Code｜🌐 Project Page]
• UniForward: Unified 3D Scene and Semantic Field Reconstruction via Feed-Forward Gaussian Splatting from Only Sparse-View Images. [📄 Paper]
• Large Spatial Model: End-to-end Unposed Images to Semantic 3D. [📄 Paper | 💻 Code｜🌐 Project Page]
• AlignGS: Aligning Geometry and Semantics for Robust Indoor Reconstruction from Sparse Views. [📄 Paper｜🌐 Project Page]

3D Scene Understanding

• MLLMs Need 3D-Aware Representation Supervision for Scene Understanding. [📄 Paper | 💻 Code｜🌐 Project Page]
• Spatio-Temporal LLM: Reasoning about Environments and Actions. [📄 Paper | 💻 Code｜🌐 Project Page]
• Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence. [📄 Paper | 💻 Code｜🌐 Project Page]
• Learning from Videos for 3D World: Enhancing MLLMs with 3D Vision Geometry Priors. [📄 Paper | 💻 Code｜🌐 Project Page]
• VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstruction. [📄 Paper | 💻 Code｜🌐 Project Page]

Video Generation

Reconstruction-enhanced Video Generation

• MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views. [📄 Paper | 💻 Code｜🌐 Project Page]
• JOG3R: Towards 3D-Consistent Video Generators. [📄 Paper｜🌐 Project Page]
• GenFusion: Closing the Loop between Reconstruction and Generation via Videos. [📄 Paper | 💻 Code｜🌐 Project Page]
• ReVision: High-Quality, Low-Cost Video Generation with Explicit 3D Physics Modeling for Complex Motion and Interaction. [📄 Paper｜🌐 Project Page]

Video Generation-based Scene Reconstruction

• Geometry Forcing: Marrying Video Diffusion and 3D Representation for Consistent World Modeling. [📄 Paper | 💻 Code｜🌐 Project Page]
• Video World Models with Long-term Spatial Memory. [📄 Paper｜🌐 Project Page]
• SteerX: Creating Any Camera-Free 3D and 4D Scenes with Geometric Steering. [📄 Paper | 💻 Code｜🌐 Project Page]
• 4DNeX: Feed-Forward 4D Generative Modeling Made Easy. [📄 Paper | 💻 Code｜🌐 Project Page]
• Lyra: Generative 3D Scene Reconstruction via Video Diffusion Model Self-Distillation. [📄 Paper]
• ShapeGen4D: Towards High Quality 4D Shape Generation from Videos. [📄 Paper｜🌐 Project Page]
• EvoWorld: Evolving Panoramic World Generation with Explicit 3D Memory. [📄 Paper]
• WorldForge: Unlocking Emergent 3D/4D Generation in Video Diffusion Model via Training-Free Guidance. [📄 Paper]
• FantasyWorld: Geometry-Consistent World Modeling via Unified Video and 3D Prediction. [📄 Paper]

Others

Panorama

• Splatter-360: Generalizable 360° Gaussian Splatting for Wide-baseline Panoramic Images. [📄 Paper | 💻 Code｜🌐 Project Page]
• PanSplat: 4K Panorama Synthesis with Feed-Forward Gaussian Splatting. [📄 Paper | 💻 Code｜🌐 Project Page]
• PanoVGGT: Feed-Forward 3D Reconstruction from Panoramic Imagery. [📄 Paper]

Localization

• Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization. [📄 Paper]
• A Scene is Worth a Thousand Features: Feed-Forward Camera Localization from a Collection of Image Features. [📄 Paper]
• Multi-View 3D Point Tracking. [📄 Paper｜🌐 Project Page]
• SAIL-Recon: Large SfM by Augmenting Scene Regression with Localization. [📄 Paper | 💻 Code｜🌐 Project Page]

Digital Humans

• Human3R: Everyone Everywhere All at Once. [📄 Paper | 💻 Code｜🌐 Project Page]

Calibration, Inpainting and Reflection

• LoRA3D: Low-Rank Self-Calibration of 3D Geometric Foundation Models. [📄 Paper｜🌐 Project Page]
• BevSplat: Resolving Height Ambiguity via Feature-Based Gaussian Primitives for Weakly-Supervised Cross-View Localization.
• InstaInpaint: Instant 3D-Scene Inpainting with Masked Large Reconstruction Model. [📄 Paper｜🌐 Project Page]
• Reflect3r: Single-View 3D Stereo Reconstruction Aided by Mirror Reflections. [📄 Paper｜🌐 Project Page]