3D generation: research map
2,448 papers on 3D generation in 3D vision, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 4 clusters and 16 approaches. The busiest year so far is 2026.
What it is
Creating and understanding 3D content: generating objects and scenes from text or images (for example with score distillation from 2D diffusion models), completing 3D scenes for driving, generating human motion in scenes and understanding indoor spaces.
Terms you will see on this page:
- Text-to-3D
- generating a 3D object from a text prompt.
- Score distillation (SDS)
- optimising a 3D model so its renders look right to a 2D diffusion model.
- Occupancy prediction
- predicting which 3D cells around a car are filled.
Start reading with the most cited papers here:
- ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes (CVPR 2017 · 4,275 citations)
- Scalability in Perception for Autonomous Driving: Waymo Open Dataset (CVPR 2020 · 3,252 citations)
Within 3D vision, its share grew from 25.4% in 2023–24 to 31.9% in 2025–26 (453 → 881 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore 3D generation in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Diffusion-based novel view synthesis rising: 8% → 13% of the topic (37 → 112 papers)
- Learning physics from video rising: 2% → 6% of the topic (11 → 54 papers)
- Feed-forward 3D asset generation and editing rising: 7% → 9% of the topic (31 → 81 papers)
- Learning 3D reconstruction from images falling: 15% → 9% of the topic (70 → 79 papers)
- Score distillation for text-to-3D falling: 11% → 6% of the topic (50 → 55 papers)
Papers from 2025–2026 that are already cited most:
- Structured 3D Latents for Scalable and Versatile 3D Generation (CVPR 2025 · 247 citations)
- Continuous 3D Perception Model with Persistent State (CVPR 2025 · 81 citations)
- VISTA3D: A Unified Segmentation Foundation Model For 3D Medical Imaging (CVPR 2025 · 69 citations)
- GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (CVPR 2025 · 49 citations)
Approaches and key papers
3D scene completion and occupancy · 824 papers
Keywords: semantic, occupancy, driving
Approaches in this cluster:
- Learning 3D reconstruction from images (274 papers)
Reconstructs 3D structure from single or multiple views with learned priors and foundation-model features. - Semantic scene completion (236 papers)
Predicts complete semantic 3D scenes from monocular images using foundation-model guidance and scene representations. - 2D-to-3D pretrained representations (108 papers)
Transfers pretrained 2D models to 3D, medical and molecular representation learning. - 3D perception for autonomous driving (119 papers)
Targets lanes, mapping, pretraining and rendering for driving, with large benchmarks. - Vision-based 3D occupancy prediction (87 papers)
Predicts semantic occupancy with efficient representations such as octrees, superquadrics and low-rank decomposition.
Most cited and most cited since 2024:
- Scalability in Perception for Autonomous Driving: Waymo Open Dataset (CVPR 2020 · 3,252 citations)
- Volumetric and Multi-View CNNs for Object Classification on 3D Data (CVPR 2016 · 1,610 citations)
- Reliable Conflictive Multi-View Learning (AAAI 2024 · 126 citations)
- Continuous 3D Perception Model with Persistent State (CVPR 2025 · 81 citations)
Text- and image-to-3D generation · 732 papers
Keywords: 3d generation, text, consistent
Approaches in this cluster:
- Diffusion-based novel view synthesis (264 papers)
Generates consistent novel views from one or few images using multi-view and video diffusion priors. - Score distillation for text-to-3D (181 papers)
Distills 2D diffusion priors into 3D with multi-view consistency to produce text-conditioned assets. - Feed-forward 3D asset generation and editing (181 papers)
Generates, textures and edits 3D content using reconstruction models and decomposed outputs. - Text-to-3D scene generation (106 papers)
Creates 3D scenes from text via latent diffusion, direct training and compositional factoring.
Most cited and most cited since 2024:
- Zero-Shot Text-Guided Object Generation With Dream Fields (CVPR 2022 · 381 citations)
- Wonder3D: Single Image to 3D using Cross-Domain Diffusion (CVPR 2024 · 341 citations)
- Structured 3D Latents for Scalable and Versatile 3D Generation (CVPR 2025 · 247 citations)
- One-2-3-45++: Fast Single Image to 3D Objects with Consistent Multi-View Generation and 3D Diffusion (CVPR 2024 · 141 citations)
Human motion and interaction · 548 papers
Keywords: motion, interaction, human
Approaches in this cluster:
- Monocular 3D object reconstruction (135 papers)
Reconstructs objects and dynamic scenes from monocular input using synthetic data and pose-free online methods. - Hand-object interaction generation (183 papers)
Generates 3D hand-object and grasp motion with contact guidance and diffusion models. - Scene-aware human motion synthesis (133 papers)
Generates or predicts human motion conditioned on 3D scenes, text and gaze. - Learning physics from video (97 papers)
Learns simulators and physical properties from videos, often with Gaussian-based scene representations.
Most cited and most cited since 2024:
- Hand Keypoint Detection in Single Images Using Multiview Bootstrapping (CVPR 2017 · 1,182 citations)
- Generating Diverse and Natural 3D Human Motions From Text (CVPR 2022 · 553 citations)
- EMAGE: Towards Unified Holistic Co-Speech Gesture Generation via Expressive Masked Audio Gesture Modeling (CVPR 2024 · 79 citations)
- Editable Scene Simulation for Autonomous Driving via Collaborative LLM-Agents (CVPR 2024 · 73 citations)
Indoor scene generation and understanding · 344 papers
Keywords: layout, indoor, 3d scene
Approaches in this cluster:
- Generative indoor scene synthesis (190 papers)
Generates 3D indoor layouts using scene graphs, diffusion, grammars and language-driven spatial reasoning. - Open-vocabulary 3D scene understanding (136 papers)
Aligns 3D scenes with language for open-world, part-level and affordance understanding. - Visual-acoustic scene modeling (18 papers)
Synthesizes room acoustics and novel-view sound using simulation and learned acoustic fields.
Most cited and most cited since 2024:
- ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes (CVPR 2017 · 4,275 citations)
- OpenScene: 3D Scene Understanding With Open Vocabularies (CVPR 2023 · 365 citations)
- DiffuScene: Denoising Diffusion Models for Generative Indoor Scene Synthesis (CVPR 2024 · 93 citations)
- Holodeck: Language Guided Generation of 3D Embodied AI Environments (CVPR 2024 · 91 citations)
Related topics in 3D vision
- Gaussian splatting (1,340)
- Human pose estimation (2,063)
- NeRF and radiance fields (733)
- Face and shape (2,246)
- Depth and stereo (2,574)
