Pith. sign in

REVIEW 4 major objections 5 minor 37 references

Toward Scene Graph and Layout Guided Complex 3D Scene Generation

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read GraLa3D generates complex 3D scenes from text by separating spatial layout from object interactions, with tests showing better prompt alignment than layout-only and scene-graph-only baselines.

desk verdict The super-node construction is genuinely new and the ablations support it, but the paper's central outperformance claim is built on an unfair comparison: GraphDreamer was run on reduced prompts. read the letter →

arxiv 2412.20473 v1 pith:GQZVQREA submitted 2024-12-29 cs.CV

Yu-Hsiang Huang , Wei Wang , Sheng-Yu Huang , Yu-Chiang Frank Wang This is my paper · ORCID
classification cs.CV
keywords text-to-3Dgenerationscenegraph3DGaussiansplattinglayout-guidedscoredistillationsamplingobjectinteractionattentionlocalization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes GraLa3D, a pipeline that turns a text description of a complex 3D scene into a structured scene graph plus layout bounding boxes. Its key claim is that splitting the graph into single-object nodes and composite super-nodes lets the system handle spatial arrangement and object interactions separately, so relations like riding, holding, or sitting on survive generation. The method optimizes each object within its layout box, jointly optimizes interacting objects through an interaction loss, and uses a localization loss to keep the objects from merging into each other. The paper reports quantitative and user-study results supporting that this approach aligns generated scenes with prompts better than either layout-only or scene-graph-only alternatives.

What carries the argument

The load-bearing object is the scene graph G = (V, E), split into single-object nodes O and super-nodes S. For each super-node, a union layout box and an interaction loss optimize paired 3D Gaussian splats together, while a localization loss matches each object's rendered alpha mask to its cross-attention map, preventing one object's Gaussians from being absorbed into the other. A masked interval-score-matching loss restricts diffusion guidance to the projected layout box, and a final harmonization stage refines the extracted mesh textures with depth-conditioned global guidance.

What would settle it

Take a fixed set of prompts whose verbs require physical contact (holding, sitting on, riding), inspect the LLM layout boxes before generation, and render the final scenes: if a substantial share of contact pairs have non-overlapping boxes and the renders show objects that do not touch, the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that a scene graph can be decomposed into two node types that are optimized differently: single-object nodes handle objects related only by spatial arrangement, while super-nodes bundle an object-relation-object triplet and optimize the union of their Gaussian splats with a joint interaction loss. A localization loss built from cross-attention maps keeps the two objects inside a super-node from borrowing each other's geometry and appearance, and a masked interval-score-matching loss keeps the diffusion prior from generating incomplete objects that merely fill their layout box. The result, the paper reports, is a 3D scene that respects both the layout and the interactions, on prompts with up to fifteen objects.

Load-bearing premise

The whole pipeline assumes the LLM's predicted bounding boxes place interacting objects in contact; if the boxes do not overlap for a relation like holding, neither the interaction loss nor the localization loss can fix the contact, and the scene silently loses that relation.

Editorial extensions

If this is right

  • Scenes with up to fifteen objects and five interacting groups can be generated, far beyond the roughly five objects handled by graph-only baselines.
  • Contact interactions such as riding, holding, sitting on, and hatching from can survive generation instead of producing floating or detached objects.
  • Layout boxes carry the spatial-relation burden, so interaction modeling does not have to compensate for missing spatial reasoning in the diffusion prior.
  • The final harmonization stage fuses separately generated objects into a single coherent mesh scene rather than a loose collection of assets.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the pipeline delegates spatial correctness to LLM-predicted boxes, the method's ceiling is set by the LLM's ability to place interacting objects in contact; a layout-validation step that checks overlap before generation would harden the approach.
  • The node/super-node decomposition could transfer to 2D multi-object image generation, where attention leakage between interacting instances is a known failure mode.
  • The localization loss treats cross-attention maps as pseudo-labels, so improvements in diffusion attention interpretability would directly improve disentanglement without retraining the pipeline.
  • Scaling the method to even larger scenes would mainly require better LLM layout reasoning, since spatial relations are outsourced to boxes rather than learned by the diffusion prior.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes GraLa3D, a text-to-3D framework that combines LLM-generated scene graphs with layout bounding boxes for complex multi-object 3D scene generation. The scene graph is decomposed into single-object nodes and composite super-nodes, where super-nodes are dedicated to objects involved in interactions. The method uses 3D Gaussian Splatting with masked Interval Score Matching, a layout constraint, an interaction loss over super-node object unions, and a localization loss based on cross-attention maps to prevent appearance leakage between interacting objects. A final scene harmonization stage converts the Gaussians to meshes and refines textures with a ControlNet-based prompt. The paper claims state-of-the-art performance relative to GraphDreamer and GALA3D, supported by CLIP scores, qualitative examples, a user study, and ablations of the two proposed losses.

Significance. If the method works as claimed, the contribution is practically useful: it is one of the few frameworks that explicitly address both spatial layout and inter-object interactions in text-to-3D scene generation, and the super-node decomposition is a sensible way to scale scene-graph-based generation to scenes with many objects. The two loss components are well motivated, and the ablations in §4.3 and Figures 8–9 give credible evidence that the localization loss and the masked ISM loss affect the output in the intended direction. The paper is also honest about known limitations in the supplementary material. However, the central comparative claim — that GraLa3D outperforms existing systems on complex prompts — rests on a quantitative evaluation with five scenes and no reported variance, and on a comparison in which the GraphDreamer baseline is run on reduced prompts for three of those scenes. The user study inherits the same prompt mismatch. These issues are fixable but currently prevent the paper's headline claim from being established.

major comments (4)
  1. [§4.2 and Table 1] The comparison with GraphDreamer is not like-for-like. Section 4.2 states that for Figures 1, 6, and 7 the authors 'reduce the number of objects in the text prompt when reproducing their results,' and Figure 6's caption confirms that GraphDreamer received only 'mermaid sits on a coral throne' rather than the full five-object prompt. Table 1 nevertheless reports CLIP scores for GraphDreamer on those cases and instructs the reader to refer to each figure for the text prompts used. The reported GraphDreamer scores are therefore likely computed against shorter prompts, and the headline claim that GraLa3D 'outperforms' GraphDreamer is not established. The authors should run all baselines on identical full prompts for all scenes and report the exact prompt used per method, or clearly separate the reduced-prompt results from the full-prompt comparison.
  2. [Table 1 and Supplementary B.2] The quantitative evidence is too thin to support the central claim. CLIP scores are averaged over 200 random views but are reported for only five scenes, with no standard deviation, no multiple seeds, and no statistical significance testing. CLIP score is also largely insensitive to spatial arrangement and interaction correctness, which are the paper's main contributions. The user study in Supplementary B.2 has the same prompt-mismatch problem: raters compared outputs that were generated under different prompts for GraphDreamer, so the preference percentages (e.g., 89.7% for GraLa3D on the farm case) may reflect the difficulty of the prompt assigned to the baseline rather than the quality of the method. A corrected evaluation should use identical full prompts for every method and report per-method prompts, variances, and preferably an interaction-aware metric such as mask-overlap or relation classification.
  3. [§3.2 and Supplementary B.3] The method's reliance on LLM-generated bounding boxes is a load-bearing assumption that the paper itself flags but does not quantify. Section 3.2 uses the LLM's bounding boxes to define both spatial relations and the union regions for interaction losses, yet Supplementary B.3 reports that for 'a monkey holding a plate' the LLM sometimes generates non-overlapping boxes, requiring manual prompt adjustment to fix the layout. If such manual adjustments were used in the main figures, they should be reported; otherwise the automated pipeline may not reproduce the shown results. The authors should report the success rate of the automatic layout generation over a larger prompt set and include a failure analysis showing how often manual intervention is needed.
  4. [§3.3] The localization loss in Eq. (5)–(6) uses cross-attention maps from the diffusion model as pseudo-ground-truth masks. This is a reasonable self-supervised bootstrap rather than a circular derivation, but the paper does not discuss its failure modes: DAAM-style attention maps can be coarse, mislocalized, or multi-modal, especially for small objects or abstract relations. The ablation in Figure 8 shows only two examples, which is insufficient to demonstrate that the loss robustly prevents appearance leakage across object classes and scene configurations. Adding quantitative results for the localization loss — e.g., mask IoU against manually annotated regions, or the fraction of super-node generations that exhibit leakage — would strengthen the central claim.
minor comments (5)
  1. [§3.3] There is a typo in the sentence introducing Eq. (7): 'upernode' should be 'super-node.'
  2. [Figure 9 caption] The caption uses 'Lmsk_ISM' while the main text and Eq. (2) use 'L_mask_ISM'; please unify the notation.
  3. [Figure 5 caption] 'wizard-crustal ball pair' should be 'wizard-crystal ball pair.'
  4. [§2.2] The sentence 'GraphDreamer [6] proposed another research direction employing scene graphs' is followed by a grammar issue in 'where each model objects that are involved'; please rephrase for clarity.
  5. [Supplementary A.2] The cross-attention extraction is described only in the supplementary material and the main text refers to DAAM without citing it at the point of use; please add the citation in §3.3 where L_local is introduced.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: GraLa3D's losses are optimization objectives over external diffusion priors, and no fitted parameter is relabeled as a prediction.

full rationale

The paper contains no derivation that reduces to its own inputs. Each loss term is an optimization objective: the masked ISM loss and the layout loss are computed from LLM-produced bounding boxes and external diffusion priors (MVDream, Interval Score Matching, ControlNet), while the localization loss aligns a 3D object's alpha mask to cross-attention maps of a pretrained diffusion model. That is a self-supervised bootstrap used to drive generation, not a fitted parameter presented as a prediction. The super-node construction is an architectural design choice, not a quantity derived from the method's own outputs. I find no load-bearing self-citation: the cited priors and baselines (DreamFusion, LucidDreamer, MVDream, GraphDreamer, GALA3D) are external works with no author overlap with the present paper. The headline comparison is weakened by the unequal prompt reduction for GraphDreamer in Figures 1, 6, and 7 (Section 4.2), but that is an evaluation-fairness issue rather than circularity: the method's outputs are not defined in terms of the baseline scores, and the fairness flaw does not make any claimed derivation equivalent to its inputs. Therefore no circular step can be exhibited, and the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the correctness of LLM-generated scene structures, on 2D diffusion priors as optimization guides, and on attention maps as instance localization. These are plausible but not guaranteed, and the paper itself documents failures in layout generation.

free parameters (3)
  • LLM-generated layout bounding boxes B = per-prompt, e.g. box coordinates for each object
    All spatial constraints and interaction modeling depend on these boxes; when they are wrong the output fails (Supp. B.3).
  • Localization loss warmup schedule = L_local weight set to 0 for first 600 iterations
    Schedules the attention-based constraint; chosen by hand in Supp. A.3 and affects training stability.
  • CLIP score evaluation sampling = 200 random views per scene, no variance
    Quantitative comparison depends on this sampling; without repeated seeds or confidence intervals the reported numbers are hard to interpret.
assumptions (4)
  • domain assumption Pre-trained 2D diffusion models (MVDream, ControlNet, Stable Diffusion) provide valid gradients for 3D optimization via ISM.
    Invoked throughout Sec. 3.3; SDS/ISM is known to have artifacts, and the method inherits this risk.
  • domain assumption DAAM cross-attention maps reliably localize each object instance in a rendered 2D view.
    Used as pseudo ground truth for L_local in Sec. 3.3 and Supp. A.2; noisy attention maps would mislead localization.
  • domain assumption LLM-generated scene graphs and layout bounding boxes are physically plausible and consistent with the text prompt.
    Used in Sec. 3.2; Supp. B.3 admits inconsistency for 'holding' and requires manual correction.
  • domain assumption Occupancy extraction from 3DGS followed by marching cubes yields faithful mesh geometry.
    Used in the harmonization stage after DreamGaussian; Supp. B.3 states mesh geometry can be inaccurate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Toward Scene Graph and Layout Guided Complex 3D Scene Generation." pith.science (2026). https://pith.science/paper/GQZVQREA

@misc{pith2026241220473,
  author       = {Pith},
  title        = {Pith review of: Toward Scene Graph and Layout Guided Complex 3D Scene Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GQZVQREA}},
  note         = {Machine review of arXiv:2412.20473}
}
read the original abstract

Recent advancements in object-centric text-to-3D generation have shown impressive results. However, generating complex 3D scenes remains an open challenge due to the intricate relations between objects. Moreover, existing methods are largely based on score distillation sampling (SDS), which constrains the ability to manipulate multiobjects with specific interactions. Addressing these critical yet underexplored issues, we present a novel framework of Scene Graph and Layout Guided 3D Scene Generation (GraLa3D). Given a text prompt describing a complex 3D scene, GraLa3D utilizes LLM to model the scene using a scene graph representation with layout bounding box information. GraLa3D uniquely constructs the scene graph with single-object nodes and composite super-nodes. In addition to constraining 3D generation within the desirable layout, a major contribution lies in the modeling of interactions between objects in a super-node, while alleviating appearance leakage across objects within such nodes. Our experiments confirm that GraLa3D overcomes the above limitations and generates complex 3D scenes closely aligned with text prompts.

Figures

Figures reproduced from arXiv: 2412.20473 by the authors.

Figure 1
Figure 1. Illustration of complex 3D scene generation. Existing text-to-3D methods generally model the layout between objects (e.g., [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Method Overview of GraLa3D. The proposed method consists of three stages: (a) Scene Graph Composition, (b) Node-to-3D Generation, and (c) 3D Scene Harmonization. Stage (a) converts a text prompt y^g into a scene graph G representation with the associated layout bounding boxes B . With nodes corresponding to objects with or without interaction, Stage (b) generates 3DGS aligned with the information described in (a). F… view at source ↗
Figure 3
Figure 3. Scene graph composition. We utilize LLM to construct a scene graph describing objects and their relations. In particular, nodes in blue denotes single objects in the scene, while supernodes in orange describe objects with interactions. 3. Method 3.1. Problem Formulation and Model Overview As depicted in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Super-node Generation. Given y S 1,2 as input, 3DGS models θ S 1 , θ S 2 are initialized and optimized. In the upper branch, θ S 1 , θ S 2 are jointly optimized using y S 1,2 and b S 1 ∪ b S 2 . In the lower branch, taking the horse (y S 2 ) as an example, the attentio…
Figure 5
Figure 5. Figure 5: Example text-to-3D generation with four objects. Given the text prompt of “a Wizard in front of a Wooden Desk, gazing into a Crystal Ball perched atop the Wooden Desk, with a Stack of Ancient Spell Books perched atop the Wooden Desk”, we are able to generate a wizard-c…
Figure 6
Figure 6. Figure 6: Examples of text-to-3D generation with five objects. Given the prompt of “A mermaid sits on a coral throne, guarding a treasure chest on a stone while a sea turtle swims above her.” and “A bear playing a saxophone stands on the stage, with a bar counter adjacent to it …
Figure 7
Figure 7. Figure 7: Example text-to-3D generation with seven objects. Given the text prompt of “A kitchen scene with a kitchen island. A rabbit sits on a stack of pancakes to the left of a blue jay standing on a large basket of rainbow macarons. On the right, a baby dragon hatches from a …
Figure 9
Figure 9. Figure 9: Ablation study on L msk ISM. Without the masked ISM loss (replaced with the original ISM loss [12]), the 2D diffusion prior [27] produces an incomplete object (the horse head is cut off) to fit in the projected bounding box (denoted by the dashed rectangle). With L mas…
Figure 10
Figure 10. Figure 10: ChatGPT for scene graph generation. The result shows that ChatGPT is capable of composing the scene with main objects given. The generated scene graph serves as structured input for the LLM to predict the spatial layout of the scene. The prompt we use for generating t…
Figure 11
Figure 11. Figure 11: Examples of text-to-3D generation with six objects. Given the prompt of “A rabbit is eating a cake on a plate. The plate, along with a spoon and fork, is on the table. The spoon is to the right of the plate, and the fork is to the left.” Among the three methods, GraLa…
Figure 12
Figure 12. Figure 12: Examples of text-to-3D generation with 15 objects. Given the prompt of “A lively scene in a dining room featuring a big dining table at the center. A monkey is next to the table, wearing a birthday hat. On the table, a rabbit is eating a birthday cake, and a parrot is…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 24 canonical work pages

  1. [1]

    Re-imagine the negative prompt algorithm: Transform 2d diffusion into 3d, alleviate janus problem and beyond

    Mohammadreza Armandpour, Ali Sadeghian, Huangjie Zheng, Amir Sadeghian, and Mingyuan Zhou. Re-imagine the negative prompt algorithm: Transform 2d diffusion into 3d, alleviate janus problem and beyond. ArXiv, abs/2304.04968, 2023. 3

    arXiv 2023

  2. [2]

    Chang, Thomas A

    Angel X. Chang, Thomas A. Funkhouser, Leonidas J. Guibas, Pat Hanrahan, Qi-Xing Huang, Zimo Li, Silvio Savarese, Manolis Savva, Shuran Song, Hao Su, Jianxiong Xiao, L. Yi, and Fisher Yu. Shapenet: An information-rich 3d model repository. ArXiv, abs/1512.03012, 2015. 2

    arXiv 2015

  3. [3]

    Chang, Thomas A

    Kevin Chen, Christopher Bongsoo Choy, Manolis Savva, Angel X. Chang, Thomas A. Funkhouser, and Silvio Savarese. Text2shape: Generating shapes from natural lan- guage by learning joint embeddings. In Asian Conference on Computer Vision, 2018. 2

    work page 2018

  4. [4]

    Set-the-scene: Global-local training for generating controllable nerf scenes, 2023

    Dana Cohen-Bar, Elad Richardson, Gal Metzer, Raja Giryes, and Daniel Cohen-Or. Set-the-scene: Global-local training for generating controllable nerf scenes, 2023. 2

    2023

  5. [5]

    Shapecrafter: A recursive text-conditioned 3d shape generation model

    Rao Fu, Xiaoyu Zhan, Yiwen Chen, Daniel Ritchie, and Sri- nath Sridhar. Shapecrafter: A recursive text-conditioned 3d shape generation model. ArXiv, abs/2207.09446, 2022. 2

    arXiv 2022

  6. [6]

    Graphdreamer: Compositional 3d scene synthesis from scene graphs

    Gege Gao, Weiyang Liu, Anpei Chen, Andreas Geiger, and Bernhard Sch ¨olkopf. Graphdreamer: Compositional 3d scene synthesis from scene graphs. 2024 IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) , pages 21295–21304, 2023. 1, 2, 3, 4, 6

    work page 2024

  7. [7]

    Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering

    Antoine Gu ´edon and Vincent Lepetit. Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering. CVPR, 2024. 3

    2024

  8. [8]

    Dreamtime: An improved opti- mization strategy for diffusion-guided 3d generation

    Yukun Huang, Jianan Wang, Yukai Shi, Boshi Tang, Xian- biao Qi, and Lei Zhang. Dreamtime: An improved opti- mization strategy for diffusion-guided 3d generation. In The Twelfth International Conference on Learning Representa- tions, 2024. 6

    work page 2024

Show all 37 references
  1. [9]

    Barron, P

    Ajay Jain, Ben Mildenhall, Jonathan T. Barron, P. Abbeel, and Ben Poole. Zero-shot text-guided object generation with dream fields. 2022 IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR), pages 857–866, 2021. 2

    2022

  2. [10]

    Clip-mesh: Generating textured meshes from text using pretrained image-text models

    Nasir Mohammad Khalid, Tianhao Xie, Eugene Belilovsky, and Tiberiu Popa. Clip-mesh: Generating textured meshes from text using pretrained image-text models. SIGGRAPH Asia 2022 Conference Papers, 2022. 2

    2022

  3. [11]

    Generative ai meets 3d: A survey on text-to-3d in aigc era

    Chenghao Li, Chaoning Zhang, Atish Waghwase, Lik-Hang Lee, Franc ¸ois Rameau, Yang Yang, Sung-Ho Bae, and Choong-Seon Hong. Generative ai meets 3d: A survey on text-to-3d in aigc era. ArXiv, abs/2305.06131, 2023. 2

    2023 arXiv

  4. [12]

    Luciddreamer: Towards high-fidelity text-to-3d generation via interval score match- ing

    Yixun Liang, Xin Yang, Jiantao Lin, Haodong Li, Xiao- gang Xu, and Yingcong Chen. Luciddreamer: Towards high-fidelity text-to-3d generation via interval score match- ing. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 6517–6526, 2023. 2, 3, 4, 8

    2024

  5. [13]

    Magic3d: High-resolution text-to-3d content creation

    Chen-Hsuan Lin, Jun Gao, Luming Tang, Towaki Takikawa, Xiaohui Zeng, Xun Huang, Karsten Kreis, Sanja Fidler, Ming-Yu Liu, and Tsung-Yi Lin. Magic3d: High-resolution text-to-3d content creation. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages ...

    2023

  6. [14]

    Lu, Xiaodong Lin, Hui Xiong, and Lin Wang

    Yiqi Lin, Haotian Bai, Sijia Li, H. Lu, Xiaodong Lin, Hui Xiong, and Lin Wang. Componerf: Text-guided multi-object compositional nerf with editable 3d scene layout. ArXiv, abs/2303.13843, 2023. 2, 3

    2023 arXiv

  7. [15]

    Towards implicit text-guided 3d shape generation

    Zhengzhe Liu, Yi Wang, Xiaojuan Qi, and Chi-Wing Fu. Towards implicit text-guided 3d shape generation. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 17875–17885, 2022. 2

    2022

  8. [16]

    Repaint: Inpainting using denoising diffusion probabilistic models

    Andreas Lugmayr, Martin Danelljan, Andres Romero, Fisher Yu, Radu Timofte, and Luc Van Gool. Repaint: Inpainting using denoising diffusion probabilistic models. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 11461–11471, 2022. 4

    2022

  9. [17]

    Wan-Duo Kurt Ma, J. P. Lewis, W. Kleijn, and Thomas Le- ung. Directed diffusion: Direct control of object placement through attention guidance. In AAAI Conference on Artificial Intelligence, 2023. 2, 3

    2023

  10. [18]

    Latent-nerf for shape-guided gener- ation of 3d shapes and textures

    Gal Metzer, Elad Richardson, Or Patashnik, Raja Giryes, and Daniel Cohen-Or. Latent-nerf for shape-guided gener- ation of 3d shapes and textures. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12663–12673, 2022. 1

    2023

  11. [19]

    Gpt-4 technical report, 2024

    OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, and et al. Gpt-4 technical report, 2024. 2, 4, 6

    2024

  12. [20]

    Compositional 3d scene generation using locally conditioned diffusion

    Ryan Po and Gordon Wetzstein. Compositional 3d scene generation using locally conditioned diffusion. 2024 Inter- national Conference on 3D Vision (3DV) , pages 651–663,

    2024

  13. [21]

    Barron, and Ben Milden- hall

    Ben Poole, Ajay Jain, Jonathan T. Barron, and Ben Milden- hall. Dreamfusion: Text-to-3d using 2d diffusion. ArXiv, abs/2209.14988, 2022. 1, 2, 4

    2022 arXiv

  14. [22]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In International Con...

    2021

  15. [23]

    Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer

    Robin Rombach, A. Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 10674–10685, 2021. 2

    2022

  16. [24]

    Denton, Seyed Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, Seyedeh Sara Mah- davi, Raphael Gontijo Lopes, Tim Salimans, Jonathan Ho, David J

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L. Denton, Seyed Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, Seyedeh Sara Mah- davi, Raphael Gontijo Lopes, Tim Salimans, Jonathan Ho, David J. Fleet, and Mohammad Norouzi. Photorealistic text- to-image...

    2022 arXiv

  17. [25]

    Lambourne, Ye Wang, Chin- Yi Cheng, and Marco Fumero

    Aditya Sanghi, Hang Chu, J. Lambourne, Ye Wang, Chin- Yi Cheng, and Marco Fumero. Clip-forge: Towards zero- shot text-to-shape generation. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 18582–18592, 2021. 2

    2022

  18. [26]

    Aditya Sanghi, Rao Fu, Vivian Liu, Karl D. D. Willis, Hooman Shayani, Amir Hosein Khasahmadi, Srinath Srid- har, and Daniel Ritchie. Clip-sculptor: Zero-shot generation of high-fidelity and diverse shapes from natural language. 2023 IEEE/CVF Conference on Computer Vision and P...

    2023

  19. [27]

    Yichun Shi, Peng Wang, Jianglong Ye, Mai Long, Kejie Li, and X. Yang. Mvdream: Multi-view diffusion for 3d gener- ation. ArXiv, abs/2308.16512, 2023. 2, 3, 4, 6, 8

    2023 arXiv

  20. [28]

    Denois- ing diffusion implicit models, 2022

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models, 2022. 3

    2022

  21. [29]

    Dreamgaussian: Generative gaussian splatting for ef- ficient 3d content creation

    Jiaxiang Tang, Jiawei Ren, Hang Zhou, Ziwei Liu, and Gang Zeng. Dreamgaussian: Generative gaussian splatting for ef- ficient 3d content creation. InThe Twelfth International Con- ference on Learning Representations, 2024. 4, 5, 3

    2024

  22. [30]

    What the DAAM: Interpreting stable dif- fusion using cross attention

    Raphael Tang, Linqing Liu, Akshat Pandey, Zhiying Jiang, Gefei Yang, Karun Kumar, Pontus Stenetorp, Jimmy Lin, and Ferhan Ture. What the DAAM: Interpreting stable dif- fusion using cross attention. In Proceedings of the 61st An- nual Meeting of the Association for Computationa...

    2023

  23. [31]

    Yeh, and Gregory Shakhnarovich

    Haochen Wang, Xiaodan Du, Jiahao Li, Raymond A. Yeh, and Gregory Shakhnarovich. Score jacobian chaining: Lift- ing pretrained 2d diffusion models for 3d generation. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12619–12629, 2022. 1

    2023

  24. [32]

    Luciddreaming: Controllable object-centric 3d generation, 2024

    Zhaoning Wang, Ming Li, and Chen Chen. Luciddreaming: Controllable object-centric 3d generation, 2024. 2

    2024

  25. [33]

    Improving compositional text-to-image generation with large vision-language models,

    Song Wen, Guian Fang, Renrui Zhang, Peng Gao, Hao Dong, and Dimitris Metaxas. Improving compositional text-to-image generation with large vision-language models,

  26. [34]

    Dream3d: Zero-shot text-to-3d synthesis using 3d shape prior and text-to-image diffusion models

    Jiale Xu, Xintao Wang, Weihao Cheng, Yan-Pei Cao, Ying Shan, Xiaohu Qie, and Shenghua Gao. Dream3d: Zero-shot text-to-3d synthesis using 3d shape prior and text-to-image diffusion models. 2023 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 20908...

    2023

  27. [35]

    Adding conditional control to text-to-image diffusion models, 2023

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models, 2023. 3, 5, 6

    2023

  28. [36]

    A survey of diffusion based im- age generation models: Issues and their solutions

    Tianyi Zhang, Zheng Wang, Jin Huang, Mohiuddin Muham- mad Tasnim, and Wei Shi. A survey of diffusion based im- age generation models: Issues and their solutions. ArXiv, abs/2308.13142, 2023. 2, 3

    2023 arXiv

  29. [37]

    A rabbit is eating a cake on a plate. The plate, along with a spoon and fork, is on the table. The spoon is to the right of the plate, and the fork is to the left

    Xiaoyu Zhou, Xingjian Ran, Yajiao Xiong, Jinlin He, Zhi- wei Lin, Yongtao Wang, Deqing Sun, and Ming-Hsuan Yang. Gala3d: Towards text-to-3d complex scene genera- tion via layout-guided generative gaussian splatting. ArXiv, abs/2402.07207, 2024. 1, 2, 3, 4, 6 Toward Scene Graph...

    2024 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.