Pith. sign in

REVIEW 5 major objections 6 minor 58 references

Chirpy3D: Part-Aware Multi-View Diffusion for Creative Fine-Grained Object Generation

T0 review · 5 major / 6 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 Chirpy3D generates never-before-seen 3D bird species by sampling and mixing learned part codes from unposed 2D photos.

desk verdict Chirpy3D is a genuine but modest step toward part-aware multi-view 3D generation; its central claim about zero-shot novel species needs stronger evidence than the paper provides. read the letter →

arxiv 2501.04144 v3 pith:6ZWFJY7P submitted 2025-01-07 cs.CV cs.GR

Kam Woh Ng , Jing Yang , Jia Wei Sii , Chee Seng Chan , Jiankang Deng , Yi-Zhe Song , Tao Xiang , Xiatian Zhu This is my paper · ORCID
classification cs.CVcs.GR
keywords fine-grained3Dgenerationmulti-viewdiffusionpart-awarepartlatentspacezero-shotcompositioncreativeobjectscoredistillationsamplingunposed2Dimages
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 tries to establish a new capability: generating entirely new, never-before-seen 3D objects (the testbed is birds) that still carry species-specific fine-grained details, using only unposed 2D images of already-seen species as training data. It argues that a hierarchical part latent space, learned without 3D data, camera poses, or manual part annotations, is enough to split an object into reusable parts, sample or interpolate those parts, and condition a multi-view diffusion model to produce coherent views of the resulting hybrid. If this works, creative 3D asset generation for games, film, or design can start from ordinary images rather than expensive 3D scans or manual part labels. The paper reports that the resulting system surpasses existing part-composition and subject-fidelity baselines on fine-grained bird data, and that the same recipe transfers to dogs, quadrupeds, and character faces.

What carries the argument

The load-bearing object is the hierarchical part latent space: each species has an embedding which a small network maps to M part codes; each part code is concatenated with a part-specific positional embedding shared across all species, then projected into the text-embedding space of a pretrained multi-view diffusion model. A Gaussian regularization loss makes the part space continuous, so users can select parts from seen species, interpolate between them, or sample novel parts from the learned distribution. Two auxiliary losses keep the machinery honest: an entropy-based attention loss that forces each part token to attend to its own spatial region, and a feature-consistency loss that minimizes cross-attention feature differences across noise levels, stabilizing denoising for unseen part combinations.

What would settle it

Take a trained Chirpy3D model and draw one random part latent from the Gaussian prior; generate the same unseen hybrid at many camera poses and from multiple seeds. If the species-defining parts (such as beak and wing colors) or the overall silhouette change identity across nearby viewpoints, or if face duplication appears, the central claim of coherent unseen-part generation is refuted. A second check is to compare part-composition success on species pairs far from the training distribution; if the attention loss cannot localize a transferred part, the shared positional-embedding alignment claim fails.

Watch

Extended reading notes

Core claim

The central claim is that a multi-view diffusion model can be made part-aware by learning a continuous latent distribution for each object part, regularized toward a standard Gaussian, and conditioning the model on part-level textual embeddings derived from that latent space. This yields, for the first time, a system that generates entirely novel 3D objects with species-specific fine-grained details through flexible part sampling and composition. The model learns the part layout implicitly from off-the-shelf 2D segmentation masks via an attention loss, and a self-supervised feature-consistency loss enforces cross-view structural and semantic coherence even for parts never seen in training. Downstream 3D reconstruction via score distillation is treated as an orthogonal step that consumes the generated multi-view images.

Load-bearing premise

The whole pipeline assumes that the consistency prior learned by the pretrained multi-view diffusion model survives two distortions: being fine-tuned with part-aware prompts, and being asked to generate unseen part combinations that were never in its training distribution; the paper's own limitation section concedes that generalizability is capped by the base model's multi-view consistency, lighting, and pose control.

Editorial extensions

If this is right

  • Bird species are a proof of concept: the paper shows the same part-space recipe interpolating and sampling on dog breeds, quadruped categories, and character faces, so the method generalizes to any fine-grained visual category with recurring parts.
  • Part-level control becomes a prompt-level operation: swapping a head, wing, or tail is done by swapping the corresponding latent code, giving artists an editing interface that needs no segmentation or 3D tools at inference time.
  • Because generated multi-view images are consistent, score-distillation 3D generation can run at lower guidance scales, reducing oversaturation artifacts while keeping the object's identity stable.
  • The learned part latent space can also be inverted on an input image, enabling reconstruction and part-level editing of real photographs without part annotations.

Reading between the lines

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

  • A likely near-term application is a data flywheel: because the framework generates coherent multi-view images of novel hybrids without 3D ground truth, those images could serve as pseudo-training data for an image-to-multi-view model, closing the loop the paper notes as future work.
  • The Gaussian part prior suggests a principled way to measure part novelty: the distance of a sampled part code from the training distribution could predict when the base model's consistency prior will break, and could be surfaced as a confidence score to users.
  • The paper's own caveat that each part code mixes structure and texture means part-level swapping will also mix appearance; separating those two factors would be the natural next axis of control.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. Chirpy3D proposes a part-aware multi-view diffusion framework for fine-grained 3D object generation from unposed 2D images, without 3D supervision or camera poses. The method learns a hierarchical part latent space (species-level embeddings mapped to part embeddings), regularizes the part latents toward a Gaussian, and projects them into textual tokens that condition a fine-tuned MVDream backbone. A self-supervised feature consistency loss L_cl is added to improve coherence for unseen part latents, and downstream 3D generation is performed via SDS with NeRF or 3DGS. Experiments on CUB-200-2011 compare against Textual Inversion and PartCraft using subject fidelity, part composition, diversity, and qualitative 3D results, with an additional user study for L_cl and extensions to dogs, quadrupeds, and faces.

Significance. If the central claims hold, the paper offers a practical recipe for creative fine-grained 3D generation from unposed 2D images, and the combination of a hierarchical part latent space with a pretrained multi-view diffusion model is a sensible and timely idea. The release of code, the inclusion of a user study, and generalization experiments across several categories are strengths. However, the evidence for the key claim of coherent multi-view generation of unseen part combinations is incomplete: no quantitative cross-view consistency metric is reported, the proposed consistency loss as written does not compare features across views, and the quantitative gains over PartCraft are marginal in several tables. The contribution is potentially significant but currently under-supported.

major comments (5)
  1. [Sec. 4.1, Tables 1–3; Sec. 5, Limitation] The paper claims coherent multi-view and 3D generation for unseen part combinations, but no quantitative cross-view consistency metric is reported. The evaluations measure subject fidelity to real images (Table 1), part composition accuracy (Table 2), and diversity (Table 3); none measures agreement among the four generated views. Given that the Limitation section explicitly concedes generalizability is 'limited by constraints in the base model, particularly in multi-view consistency,' this missing evaluation is load-bearing. The authors should report pairwise view similarity (e.g., DINO/CLIP cosine similarity or LPIPS across generated views) for seen and novel part latents, and compare with the MVDream backbone and PartCraft.
  2. [Sec. 3.3, Eq. (6), Fig. 3] The self-supervised feature consistency loss L_cl is claimed to enforce 'structural and semantic coherence across views,' but the equation compares cross-attention feature maps F_ϵi and F_ϵj for two different random noise inputs at the same timestep t. This is a single-image denoising-stability term, not a cross-view term; it does not involve features from different camera views. The claim in Sec. 3.3 and the name 'feature consistency loss' are therefore overstated. Either the loss should compare features across the four views in a multi-view batch, or the paper should reframe it as a denoising-stability regularizer whose benefit to SDS comes from reduced variance across noise draws.
  3. [Sec. 3.2 vs Sec. 3.3, Eq. (1) vs Eq. (5)] The method is described as regularizing part embeddings to a standard Gaussian and enabling 'novel part synthesis via a standard Gaussian prior' (Fig. 2 caption), but the actual sampling distribution in Eq. (5) is N(µ_m, σ^2_m) with per-part empirical mean and variance computed over training species. These are not standard Gaussian, and the L2 penalty in Eq. (1) does not ensure the empirical per-part statistics used at inference. The paper should either sample from the regularized prior (N(0, I)) or justify the empirical distribution and reconcile it with the stated regularization.
  4. [Sec. 4.1, Tables 1–2] The quantitative claims of superiority are not supported by the reported numbers. In Table 1, Chirpy3D's FID (43.41) is worse than PartCraft (43.25), and in Table 2 the CoSim gain over PartCraft is 0.002; no error bars or significance tests are provided anywhere. The abstract states that 'Chirpy3D surpasses existing methods in generating creative 3D objects with higher quality and fine-grained details,' but the evidence is marginal for some metrics and absent for 3D quality. The authors should add confidence intervals, significance tests, or at least temper the claims to the specific metrics that show consistent improvement.
  5. [Sec. 4.2 and Sec. 1] The 'first system capable of generating entirely novel 3D objects' claim is not supported by any comparison against 3D-capable baselines. The only competitors are Textual Inversion and PartCraft, both 2D methods adapted to MVDream; there is no comparison to prior text-to-3D or part-aware 3D generation methods, and the 3D evaluation is purely qualitative (Fig. 10). At minimum, the authors should compare against a text-to-3D baseline (e.g., MVDream SDS with a textual part prompt) and report a quantitative 3D metric or a user study on 3D objects.
minor comments (6)
  1. [Sec. 4, Implementation; Appendix C] The main text sets the part latent dimension to D_p = 42, while the supplementary (Appendix C, Table 6) states D_l = 4 as the default and shows ablations with D_l ∈ {4, 16, 32, 64}. The notation and actual value are inconsistent; please unify the notation and state which dimension is used in the main experiments.
  2. [Abstract and Sec. 1] The method is described as 'unsupervised' and 'without manual part annotations,' but the abstract also says it uses 'off-the-shelf 2D part segmentation masks as spatial guidance.' These masks are a form of part-level supervision. Please clarify whether the part decomposition is fully unsupervised or uses pretrained segmentation masks, and adjust the wording in Sec. 1 and Related Work accordingly.
  3. [Sec. 4.1] The term 'FIDCLIP' appears without definition. Please define how FIDCLIP is computed (presumably FID in a CLIP feature space) in the evaluation section.
  4. [Table 3] Table 3 reports only PartCraft and Chirpy3D, while the surrounding text discusses Textual Inversion's artifacts. Please include Textual Inversion in the table or explicitly state why it is omitted.
  5. [Appendix A] In the derivation of Lreg, the text contains a typo, 'bmmu = 0,' which should read 'µ = 0.'
  6. [Fig. 9 caption] The claim that Chirpy3D produces 'well-clustered DINO feature representations' is supported only by a t-SNE plot; consider reporting a quantitative cluster metric (e.g., silhouette score) if this claim is retained.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the central part-sampling contribution is generative and externally evaluated, with only non-load-bearing reuse of the authors' PartCraft components.

full rationale

Walking the derivation chain, the species embeddings s_c are mapped by f to part latents p_c_m, regularized by Eq. (1), projected by g to part-aware text tokens (Eq. 2), and used to condition a fine-tuned MVDream. Every step is an optimization or sampling operation; no predicted quantity is defined as the fitted value it is claimed to predict. The Gaussian prior (Eqs. (1), (5)) is a regularizer and sampling mechanism, not a fitted prediction, and the diversity and quality evaluations are measured against real training images via DINO/CLIP retrieval or user preference, i.e., outside the fitted constants. The paper does reuse the entropy attention loss and segmentation module from the authors' PartCraft [35], and uses PartCraft as a baseline, but this is component reuse and comparison, not load-bearing circularity: the central contribution is the hierarchical part latent space and sampling/interpolation behavior, which does not reduce to PartCraft's loss. Lcl (Eq. 6) is described as enforcing cross-view coherence but as written compares feature maps of different noise inputs at the same timestep, and no direct cross-view consistency metric is reported; this is an evidence and correctness gap, not a circular reduction. The Limitation section candidly attributes generalizability limits to the base model's multi-view consistency. No step meets the quoted-reduction bar, so the paper is not significantly circular.

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

The method rests on a pretrained multi-view diffusion backbone, an off-the-shelf part segmentation module, and a chosen Gaussian regularization; it introduces no physically new entities. The free parameters are the part count, latent dimension, and loss weights, plus the LoRA rank and guidance scales.

free parameters (3)
  • Number of parts M = 5
    Set manually to head, wings, torso, legs, and tail for birds; the framework requires specifying M before training.
  • Part latent dimension D_p = 42 (main text); supplementary ablation reports D_l=4 as the chosen default for diversity
    The dimension of the part latent space is a design choice; supplementary Table 6 tunes it on the entropy/diversity metric.
  • Loss weights lambda_attn, lambda_cl, lambda_reg = 0.01, 0.001, 0.0001
    Chosen by hand; Appendix C says lambda_cl above 0.001 degrades quality, so the value is selected after observing outputs.
assumptions (4)
  • domain assumption MVDream's pretrained multi-view consistency transfers to the fine-tuned part-aware prompts and to unseen part combinations
    The whole framework relies on the backbone's multi-view prior; the paper's limitation section notes generalizability is constrained by the base model.
  • domain assumption The off-the-shelf part segmentation module from PartCraft gives reliable part masks for the attention loss on unseen images
    Equation 4 uses S_m as supervision; if masks are wrong, part disentanglement weakens.
  • ad hoc to paper Part latent codes are well modeled by a standard Gaussian distribution
    Equation 1 regularizes embeddings to N(0,I) to enable sampling and interpolation; this is a design choice, not an empirical law.
  • domain assumption Cross-attention maps in the U-Net localize the corresponding object parts after fine-tuning
    The attention loss and the feature-consistency loss both assume attention maps correlate with spatial part regions; Appendix D provides evidence but this is assumed during training.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Chirpy3D: Part-Aware Multi-View Diffusion for Creative Fine-Grained Object Generation." pith.science (2026). https://pith.science/paper/6ZWFJY7P

@misc{pith2026250104144,
  author       = {Pith},
  title        = {Pith review of: Chirpy3D: Part-Aware Multi-View Diffusion for Creative Fine-Grained Object Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6ZWFJY7P}},
  note         = {Machine review of arXiv:2501.04144}
}
read the original abstract

Understanding and generating the fine-grained structure of objects -- such as birds with species-specific beaks, wings, and tails -- is a long-standing challenge in computer vision. We propose Chirpy3D, a part-aware multi-view diffusion framework that learns a hierarchical part latent space from unposed 2D images, using only off-the-shelf 2D part segmentation masks as spatial guidance -- without requiring any 3D data, camera poses, or manual part annotations. This latent space enables intuitive part-level swapping, interpolation, and zero-shot composition. A self-supervised feature consistency loss further encourages structural alignment across views, allowing coherent generation even with hybrid or unseen part combinations. Our core contribution is the controllable part-aware latent space and multi-view diffusion model. Downstream 3D generation is supported via any differentiable renderer such as NeRF but is orthogonal to the main framework, making Chirpy3D a flexible foundation for creative object generation in the absence of structured 3D data. Code is released at https://github.com/kamwoh/chirpy3d.

Figures

Figures reproduced from arXiv: 2501.04144 by the authors.

Figure 1
Figure 1. Novel, creative species created by our Chirpy3D. Feel free to name them! [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of Chirpy3D. Chirpy3D takes (a) a set of unposed 2D images from multiple fine-grained species (e.g., birds) and (b) learns to decompose each object into a set of underlying parts (e.g., head, wings, torso, legs, tail) within a hierarchical part latent space – species embedding s captures glboal species characteristics while part-level embedding p captures fine-grained part variations. (c) A regularized part… view at source ↗
Figure 4
Figure 4. (a) Seen part selection generation. Unseen part synthesis via (b) novel sampling and (c) interpolation. different noise levels: Lcl = Et,ϵ,L   X i,j ∥Fϵi − Fϵj ∥ 2   . (6) where i, j represent different random noise inputs at the same timestep t, and L denotes different layers in the U￾Net. See [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (17 more)
Figure 5
Figure 5. Figure 5: Multi-view subject-driven generation on two species - [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Visual comparison of part composition. A, B, C, D, E, F represent cardinal, wilson warbler, least auklet, california gull, horned lark, and song sparrow respectively. Red circles indicate changed parts. All generated (including sources & targets) by the same seed. Amon…
Figure 7
Figure 7. Figure 7: Linear interpolation of all part latents between two [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Generated images with random sampled latents/embeddings. Textual Inversion often produces images with artifacts due to [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: t-SNE embeddings of DINO features of generated im [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]
Figure 10
Figure 10. Figure 10: NeRF rendering of learned 3D objects. Chirpy3D With Lcl Without Lcl Preference ↑ 82.5% 17.5% [PITH_FULL_IMAGE:figures/full_fig_p007_10.png]
Figure 11
Figure 11. Figure 11: (Top) Middle images indicate a hybrid between species/instance left and right by linearly interpolating their part latent codes. (Bottom) Generated images with random sampled latents for PartImageNet (quadruped only) [19] and sims4-faces [41] We regularize part embedd…
Figure 12
Figure 12. Figure 12: All images are generated with the same camera pose [PITH_FULL_IMAGE:figures/full_fig_p008_12.png]
Figure 13
Figure 13. Figure 13: Comparison of generated images with different scales, [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]
Figure 14
Figure 14. Figure 14: Image reconstruction from part latent inversion. We [PITH_FULL_IMAGE:figures/full_fig_p013_14.png]
Figure 16
Figure 16. Figure 16: Visualizing part latent space via t-SNE embeddings. [PITH_FULL_IMAGE:figures/full_fig_p014_16.png]
Figure 18
Figure 18. Figure 18: Image-to-3D using front view and side view of gener [PITH_FULL_IMAGE:figures/full_fig_p014_18.png]
Figure 19
Figure 19. Figure 19: Multi-view generation with text prompt through MVDream [ [PITH_FULL_IMAGE:figures/full_fig_p015_19.png]
Figure 20
Figure 20. Figure 20: Multi-view generation on existing species, trained with respective methods [PITH_FULL_IMAGE:figures/full_fig_p016_20.png]
Figure 21
Figure 21. Figure 21: Multi-view generation on novel species (random sampling), trained with respective methods. All were generated with the [PITH_FULL_IMAGE:figures/full_fig_p017_21.png]
Figure 22
Figure 22. Figure 22: Multi-view generation on novel species (interpolation) trained with respective methods. All images were generated using the [PITH_FULL_IMAGE:figures/full_fig_p018_22.png]
Figure 23
Figure 23. Figure 23: Each row is a different seed with a random sampled part latents [PITH_FULL_IMAGE:figures/full_fig_p019_23.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 48 canonical work pages

  1. [1]

    Break-a-scene: Ex- tracting multiple concepts from a single image

    Omri Avrahami, Kfir Aberman, Ohad Fried, Daniel Cohen-Or, and Dani Lischinski. Break-a-scene: Ex- tracting multiple concepts from a single image. In SIGGRAPH Asia, 2023. 1, 2

    work page 2023

  2. [2]

    Dreamavatar: Text-and-shape guided 3d human avatar generation via diffusion mod- els

    Yukang Cao, Yan-Pei Cao, Kai Han, Ying Shan, and Kwan-Yee K Wong. Dreamavatar: Text-and-shape guided 3d human avatar generation via diffusion mod- els. In CVPR, 2024. 1

    work page 2024

  3. [3]

    Emerging properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e J´egou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerging properties in self-supervised vision transformers. In ICCV, 2021. 5, 6

    work page 2021

  4. [4]

    Shapenet: An information-rich 3d model repos- itory

    Angel X Chang, Thomas Funkhouser, Leonidas Guibas, Pat Hanrahan, Qixing Huang, Zimo Li, Sil- vio Savarese, Manolis Savva, Shuran Song, Hao Su, et al. Shapenet: An information-rich 3d model repos- itory. arXiv preprint arXiv:1512.03012, 2015. 2

    arXiv 2015

  5. [5]

    The devil is in the channels: Mutual-channel loss for fine-grained image classifica- tion

    Dongliang Chang, Yifeng Ding, Jiyang Xie, Ayan Ku- mar Bhunia, Xiaoxu Li, Zhanyu Ma, Ming Wu, Jun Guo, and Yi-Zhe Song. The devil is in the channels: Mutual-channel loss for fine-grained image classifica- tion. TIP, 2020. 2

    work page 2020

  6. [6]

    Sphericgan: Semi- supervised hyper-spherical generative adversarial net- works for fine-grained image synthesis

    Tianyi Chen, Yunfei Zhang, Xiaoyang Huo, Si Wu, Yong Xu, and Hau San Wong. Sphericgan: Semi- supervised hyper-spherical generative adversarial net- works for fine-grained image synthesis. In CVPR,

    work page

  7. [7]

    To- wards creativity characterization of generative models via group-based subset scanning

    Celia Cintas, Payel Das, Brian Quanz, Girmaw Abebe Tadesse, Skyler Speakman, and Pin-Yu Chen. To- wards creativity characterization of generative models via group-based subset scanning. In IJCAI, 2022. 2

    work page 2022

  8. [8]

    Toward a neuro-inspired creative de- coder

    Payel Das, Brian Quanz, Pin-Yu Chen, Jae-wook Ahn, and Dhruv Shah. Toward a neuro-inspired creative de- coder. In IJCAI, 2020. 2

    work page 2020

Show all 58 references
  1. [9]

    Comgan: unsupervised disentangle- ment and segmentation via image composition

    Rui Ding, Kehua Guo, Xiangyuan Zhu, Zheng Wu, and Liwei Wang. Comgan: unsupervised disentangle- ment and segmentation via image composition. 2022. 2

    2022

  2. [10]

    Fine-grained visual classification via progres- sive multi-granularity training of jigsaw patches

    Ruoyi Du, Dongliang Chang, Ayan Kumar Bhu- nia, Jiyang Xie, Zhanyu Ma, Yi-Zhe Song, and Jun Guo. Fine-grained visual classification via progres- sive multi-granularity training of jigsaw patches. In ECCV, 2020. 2

    2020

  3. [11]

    Can: Creative adversarial networks generating ”art” by learning about styles and deviating from style norms

    Ahmed Elgammal, Bingchen Liu, Mohamed Elho- seiny, and Marian Mazzone. Can: Creative adversarial networks generating ”art” by learning about styles and deviating from style norms. In ICCC, 2017. 2

    2017

  4. [12]

    An image is worth one word: Personal- izing text-to-image generation using textual inversion

    Rinon Gal, Yuval Alaluf, Yuval Atzmon, Or Patash- nik, Amit H Bermano, Gal Chechik, and Daniel Cohen-Or. An image is worth one word: Personal- izing text-to-image generation using textual inversion. arXiv preprint arXiv:2208.01618, 2022. 5, 12

    2022 arXiv

  5. [13]

    Get3d: A generative model of high quality 3d textured shapes learned from images

    Jun Gao, Tianchang Shen, Zian Wang, Wenzheng Chen, Kangxue Yin, Daiqing Li, Or Litany, Zan Go- jcic, and Sanja Fidler. Get3d: A generative model of high quality 3d textured shapes learned from images. In NeurIPS, 2022. 2

    2022

  6. [14]

    Lawrence Zitnick, and Devi Parikh

    Songwei Ge, Vedanuj Goswami, C. Lawrence Zitnick, and Devi Parikh. Creative sketch generation. In ICLR,

  7. [15]

    Learning shape templates with structured implicit functions

    Kyle Genova, Forrester Cole, Daniel Vlasic, Aaron Sarna, William T Freeman, and Thomas Funkhouser. Learning shape templates with structured implicit functions. In ICCV, 2019. 2

    2019

  8. [16]

    Local deep implicit functions for 3d shape

    Kyle Genova, Forrester Cole, Avneesh Sud, Aaron Sarna, and Thomas Funkhouser. Local deep implicit functions for 3d shape. In CVPR, 2020. 2

    2020

  9. [17]

    threestudio: A unified framework for 3d content generation

    Yuan-Chen Guo, Ying-Tian Liu, Ruizhi Shao, Chris- tian Laforte, Vikram V oleti, Guan Luo, Chia-Hao Chen, Zi-Xin Zou, Chen Wang, Yan-Pei Cao, and Song-Hai Zhang. threestudio: A unified framework for 3d content generation. https://github. com/threestudio-project/threestudio ,

  10. [18]

    Headsculpt: Crafting 3d head avatars with text

    Xiao Han, Yukang Cao, Kai Han, Xiatian Zhu, Jiankang Deng, Yi-Zhe Song, Tao Xiang, and Kwan- Yee K Wong. Headsculpt: Crafting 3d head avatars with text. In Advances in Neural Information Process- ing Systems, 2024. 1

    2024

  11. [19]

    Partimagenet: A large, high-quality dataset of parts

    Ju He, Shuo Yang, Shaokang Yang, Adam Ko- rtylewski, Xiaoding Yuan, Jie-Neng Chen, Shuai Liu, Cheng Yang, and Alan Yuille. Partimagenet: A large, high-quality dataset of parts. arXiv preprint arXiv:2112.00933, 2021. 8

    2021 arXiv

  12. [20]

    Spaghetti: Editing implicit shapes through part aware generation

    Amir Hertz, Or Perel, Raja Giryes, Olga Sorkine- Hornung, and Daniel Cohen-Or. Spaghetti: Editing implicit shapes through part aware generation. TOG,

  13. [21]

    LoRA: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. In ICLR, 2022. 5, 12

    2022

  14. [22]

    Dreamtime: An improved optimization strategy for text-to-3d content creation

    Yukun Huang, Jianan Wang, Yukai Shi, Xianbiao Qi, Zheng-Jun Zha, and Lei Zhang. Dreamtime: An improved optimization strategy for text-to-3d content creation. arXiv preprint arXiv:2306.12422, 2023. 14

    2023 arXiv

  15. [23]

    Neural template: Topology-aware reconstruction and disentangled generation of 3d meshes

    Ka-Hei Hui, Ruihui Li, Jingyu Hu, and Chi-Wing Fu. Neural template: Topology-aware reconstruction and disentangled generation of 3d meshes. InCVPR, 2022. 2

    2022

  16. [24]

    Spad: Spatially aware multi-view dif- fusers

    Yash Kant, Aliaksandr Siarohin, Ziyi Wu, Michael Vasilkovsky, Guocheng Qian, Jian Ren, Riza Alp Guler, Bernard Ghanem, Sergey Tulyakov, and Igor Gilitschenski. Spad: Spatially aware multi-view dif- fusers. In CVPR, 2024. 1, 2 9

    2024

  17. [25]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM ToG, 2023. 5

    2023

  18. [26]

    Eschernet: A generative model for scalable view synthesis

    Xin Kong, Shikun Liu, Xiaoyang Lyu, Marwan Taher, Xiaojuan Qi, and Andrew J Davison. Eschernet: A generative model for scalable view synthesis. In CVPR, 2024. 2

    2024

  19. [27]

    Salad: Part-level latent diffusion for 3d shape generation and manipulation

    Juil Koo, Seungwoo Yoo, Minh Hieu Nguyen, and Minhyuk Sung. Salad: Part-level latent diffusion for 3d shape generation and manipulation. InICCV, 2023. 2

    2023

  20. [28]

    Tp2o: Creative text pair-to-object generation using balance swap- sampling

    Jun Li, Zedong Zhang, and Jian Yang. Tp2o: Creative text pair-to-object generation using balance swap- sampling. In ECCV, 2024. 1, 2

    2024

  21. [29]

    Mixnmatch: Multifactor disentangle- ment and encoding for conditional image generation

    Yuheng Li, Krishna Kumar Singh, Utkarsh Ojha, and Yong Jae Lee. Mixnmatch: Multifactor disentangle- ment and encoding for conditional image generation. In CVPR, 2020. 2

    2020

  22. [30]

    Partgan: Weakly-supervised part de- composition for image generation and segmentation

    Yuheng Li, Krishna Kumar Singh, Yang Xue, and Yong Jae Lee. Partgan: Weakly-supervised part de- composition for image generation and segmentation. In BMVC, 2021. 2

    2021

  23. [31]

    Zero-1-to-3: Zero-shot one image to 3d object

    Ruoshi Liu, Rundi Wu, Basile Van Hoorick, Pavel Tokmakov, Sergey Zakharov, and Carl V ondrick. Zero-1-to-3: Zero-shot one image to 3d object. In ICCV, 2023. 2

    2023

  24. [32]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In ICLR, 2018. 12

    2018

  25. [33]

    Nerf: Representing scenes as neural radiance fields for view synthesis

    B Mildenhall, PP Srinivasan, M Tancik, JT Barron, R Ramamoorthi, and R Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. In ECCV,

  26. [34]

    Concepthash: Interpretable fine-grained hashing via concept discovery

    Kam Woh Ng, Xiatian Zhu, Yi-Zhe Song, and Tao Xi- ang. Concepthash: Interpretable fine-grained hashing via concept discovery. In CVPRW, 2024. 2

    2024

  27. [35]

    Partcraft: Crafting creative objects by parts

    Kam Woh Ng, Xiatian Zhu, Yi-Zhe Song, and Tao Xi- ang. Partcraft: Crafting creative objects by parts. In ECCV, 2024. 1, 2, 3, 4, 5, 7, 12

    2024

  28. [36]

    Creativegan: Editing generative adver- sarial networks for creative design synthesis

    Amin Heyrani Nobari, Muhammad Fathy Rashad, and Faez Ahmed. Creativegan: Editing generative adver- sarial networks for creative design synthesis. arXiv preprint arXiv:2103.06242, 2021. 2

    2021 arXiv

  29. [37]

    Deepsdf: Learning continuous signed distance functions for shape representation

    Jeong Joon Park, Peter Florence, Julian Straub, Richard Newcombe, and Steven Lovegrove. Deepsdf: Learning continuous signed distance functions for shape representation. In CVPR, 2019. 2

    2019

  30. [38]

    Dreamfusion: Text-to-3d using 2d dif- fusion

    Ben Poole, Ajay Jain, Jonathan T Barron, and Ben Mildenhall. Dreamfusion: Text-to-3d using 2d dif- fusion. In ICLR, 2022. 5

    2022

  31. [39]

    Learning transferable visual models from nat- ural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sas- try, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from nat- ural language supervision. In ICML, 2021. 5

    2021

  32. [40]

    Conceptlab: Creative genera- tion using diffusion prior constraints

    Elad Richardson, Kfir Goldberg, Yuval Alaluf, and Daniel Cohen-Or. Conceptlab: Creative genera- tion using diffusion prior constraints. arXiv preprint arXiv:2308.02669, 2023. 1, 2

    2023 arXiv

  33. [41]

    Sims4 faces

    Joe Rocca. Sims4 faces. https : / / huggingface . co / datasets / rocca / sims4-faces, 2022. 8

    2022

  34. [42]

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. In CVPR, 2023. 5

    2023

  35. [43]

    Design: Design inspiration from generative networks

    Othman Sbai, Mohamed Elhoseiny, Antoine Bordes, Yann LeCun, and Camille Couprie. Design: Design inspiration from generative networks. In ECCVW,

  36. [44]

    Zero123++: a single image to consistent multi-view diffusion base model, 2023

    Ruoxi Shi, Hansheng Chen, Zhuoyang Zhang, Minghua Liu, Chao Xu, Xinyue Wei, Linghao Chen, Chong Zeng, and Hao Su. Zero123++: a single image to consistent multi-view diffusion base model, 2023. 2, 14

    2023

  37. [45]

    Mvdream: Multi-view diffusion for 3d generation

    Yichun Shi, Peng Wang, Jianglong Ye, Long Mai, Ke- jie Li, and Xiao Yang. Mvdream: Multi-view diffusion for 3d generation. In ICLR, 2023. 1, 2, 3, 5, 12, 15

    2023

  38. [46]

    Meshgpt: Gener- ating triangle meshes with decoder-only transformers

    Yawar Siddiqui, Antonio Alliegro, Alexey Artemov, Tatiana Tommasi, Daniele Sirigatti, Vladislav Rosov, Angela Dai, and Matthias Nießner. Meshgpt: Gener- ating triangle meshes with decoder-only transformers. In CVPR, 2024. 2

    2024

  39. [47]

    Finegan: Unsupervised hierarchical disentangle- ment for fine-grained object generation and discovery

    Krishna Kumar Singh, Utkarsh Ojha, and Yong Jae Lee. Finegan: Unsupervised hierarchical disentangle- ment for fine-grained object generation and discovery. In CVPR, pages 6490–6499, 2019. 2

    2019

  40. [48]

    Dreamgaussian: Generative gaussian splatting for efficient 3d content creation

    Jiaxiang Tang, Jiawei Ren, Hang Zhou, Ziwei Liu, and Gang Zeng. Dreamgaussian: Generative gaussian splatting for efficient 3d content creation. In ICLR,

  41. [49]

    Concept decomposition for visual ex- ploration and inspiration

    Yael Vinker, Andrey V oynov, Daniel Cohen-Or, and Ariel Shamir. Concept decomposition for visual ex- ploration and inspiration. In SIGGRAPH Asia, 2023. 2

    2023

  42. [50]

    The caltech-ucsd birds- 200-2011 dataset

    Catherine Wah, Steve Branson, Peter Welinder, Pietro Perona, and Serge Belongie. The caltech-ucsd birds- 200-2011 dataset. 2011. 5

    2011

  43. [51]

    Fine-grained image analysis with deep learning: A survey

    Xiu-Shen Wei, Yi-Zhe Song, Oisin Mac Aodha, Jianxin Wu, Yuxin Peng, Jinhui Tang, Jian Yang, and Serge Belongie. Fine-grained image analysis with deep learning: A survey. TPAMI, 2021. 2

    2021

  44. [52]

    Sagnet: Structure- 10 aware generative network for 3d-shape modeling

    Zhijie Wu, Xiang Wang, Di Lin, Dani Lischinski, Daniel Cohen-Or, and Hui Huang. Sagnet: Structure- 10 aware generative network for 3d-shape modeling. TOG, 2019. 2

    2019

  45. [53]

    Instantmesh: Efficient 3d mesh generation from a single image with sparse- view large reconstruction models

    Jiale Xu, Weihao Cheng, Yiming Gao, Xintao Wang, Shenghua Gao, and Ying Shan. Instantmesh: Efficient 3d mesh generation from a single image with sparse- view large reconstruction models. arXiv preprint arXiv:2404.07191, 2024. 7, 14

    2024 arXiv

  46. [54]

    Attngan: Fine-grained text to image generation with attentional generative adversarial networks

    Tao Xu, Pengchuan Zhang, Qiuyuan Huang, Han Zhang, Zhe Gan, Xiaolei Huang, and Xiaodong He. Attngan: Fine-grained text to image generation with attentional generative adversarial networks. In CVPR,

  47. [55]

    Sym3d: Learning symmetric triplanes for better 3d-awareness of gans

    Jing Yang, Kyle Fogarty, Fangcheng Zhong, and Cen- giz Oztireli. Sym3d: Learning symmetric triplanes for better 3d-awareness of gans. arXiv preprint arXiv:2406.06432, 2024. 2

    2024 arXiv

  48. [56]

    Gaussiandreamer: Fast gener- ation from text to 3d gaussians by bridging 2d and 3d diffusion models

    Taoran Yi, Jiemin Fang, Junjie Wang, Guanjun Wu, Lingxi Xie, Xiaopeng Zhang, Wenyu Liu, Qi Tian, and Xinggang Wang. Gaussiandreamer: Fast gener- ation from text to 3d gaussians by bridging 2d and 3d diffusion models. In CVPR, 2024. 1

    2024

  49. [57]

    Entropy_{n}:

    Heliang Zheng, Jianlong Fu, Zheng-Jun Zha, and Jiebo Luo. Looking for the devil in the details: Learning trilinear attention sampling network for fine- grained image recognition. In CVPR, 2019. 2 11 A. Derivation We use the symbolx to represent the part latent l for clarity. A...

    2019

  50. [58]

    a [part1] ... [partM ] bird

    as our framework andMVDream-threestudio as a plugin. We implement a custom prompt processor to handle the input prompt, as we replace the word embeddings using Eq.2. After tokenization and before passing the word em- beddings, we substitute the placeholder’s word embedding wit...

Pith tools

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