REVIEW 4 major objections 6 minor 55 references
VersaGen: Unleashing Versatile Visual Control for Text-to-Image Synthesis
T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash
Pith's one-line read VersaGen claims that text-to-image generation can take partial sketches — one object, several objects, a background, or none — instead of a full scene drawing.
desk verdict VersaGen's flexible object/background control framing is genuinely useful, but the multi-subject capability rests on an unspecified drawing-to-token mapping at inference. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The carrying mechanism is the cross-attention localization loop: Equation (2) computes $A_t^i(s_i,c_i)$ from the frozen model's queries for the drawing and keys for the text token; Equation (3) thresholds it with Otsu's method to get a region $R_i$ and bounding box $B_i$; Equation (4) concatenates the encoded drawings into a fused latent $z_S$; Equation (5) injects $z_S$ through a ControlNet-style trainable copy; Equations (7)--(10) define the Multimodal Conflict Resolver's token and pixel losses and the latent update $z'_\tau = z_\tau - \alpha \nabla_{z_\tau} L$; Equation (13) scales the drawing branch by a timestep-dependent weight $\gamma(t)=1-a/(1+e^{-b(t-c)})$. Together these let the model decide where each sketch goes, resolve text--sketch disagreements at one timestep, keep multiple objects from colliding, and let text dominate later fine-detail stages.
What would settle it
Give VersaGen a prompt such as 'a bear and a deer are in the rainforest' with two sketches whose identities are swapped relative to the nouns; if the generated image places the deer where the bear sketch was drawn, or blends the two, the claimed cross-attention localization and decoupling have failed. Alternatively, feed the same prompt with totally random doodles; if output quality remains high, the drawing encoder may be ignoring sketch content, which would contradict the central control claim.
Extended reading notes
Core claim
VersaGen's central claim is that object-level and scene-level visual control can be unified in one framework without retraining the underlying diffusion model. The paper builds on the observation that a frozen text-to-image model's cross-attention maps act as semantic segmenters: at a chosen denoising timestep $\tau$, the map for a noun token $c_i$ can be thresholded via Otsu's method to produce a bounding box, and each user drawing $s_i$ is relocated into that box. A trainable copy of the UNet, injected through zero convolutions, then conditions generation on the concatenated drawing latents. During inference, the Multimodal Conflict Resolver optimizes the noisy latent with token- and pixel-level losses so the drawing contour matches the cross-attention region; Multi-object Decoupling shrinks and translates overlapping boxes; Adaptive Control Strength decays the drawing's influence in later timesteps. On COCO and Sketchy, the paper reports that VersaGen surpasses Stable Diffusion, ControlNet, T2I-Adapter, UniControl, GLIGEN, and InstanceDiffusion on FID and DINO, with a user study in which 48% of participants preferred it for ease of use.
Load-bearing premise
The method assumes each sketch can be matched to the correct noun in the prompt, but the paper never says how that match is supplied for an arbitrary sketch.
Editorial extensions
If this is right
- Users can generate images from a partial drawing and let the model decide placement, so they no longer need to reconstruct the entire scene.
- One adapter on a frozen base model covers single-subject, multi-subject, background, and text-only control, removing the need for separate condition-specific models.
- The paper's ablation shows performance improving monotonically as control increases from text-only to single subject, multiple subjects, and subjects plus background.
- The same inference-time MCR can be attached to a plain ControlNet on SDXL and improves its scores without retraining, suggesting the conflict-resolution step is model-agnostic.
Reading between the lines
- The noun-to-sketch mapping the paper assumes is the real usability bottleneck: unless a system or interface supplies it, the cross-attention localization cannot run, and VersaGen's multi-subject promise depends on a matching step the paper leaves unspecified.
- If the localization really comes from the base model's cross-attention, then VersaGen inherits any blind spots of the frozen T2I model — rare objects, abstract nouns, or ambiguous prompts would degrade placement even with perfect sketches.
- A natural extension would be to let the user annotate the sketch-token link, for example by clicking a noun and a sketch, which would turn the assumed correspondence into an explicit input and likely improve reliability on arbitrary prompts.
- The edge-map-as-pseudo-sketch training choice may limit robustness to very abstract or stylized drawings; testing on sketch styles far from the COCO edge statistics could reveal how much of the reported gain is tied to that training distribution.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper presents VersaGen, a text-to-image synthesis system that extends a frozen Stable Diffusion backbone with a ControlNet-style adapter, enabling users to provide object-level sketches (single or multiple subjects), a background sketch, any combination of the two, or no visual input at all, with text-only fallback. The method automatically localizes user drawings via cross-attention maps, and three inference-time mechanisms are introduced: a Multimodal Conflict Resolver (MCR) that updates the noisy latent to align attention with drawing contours, a Multi-object Decoupling (MoD) heuristic that adjusts predicted bounding boxes to reduce overlap, and an Adaptive Control Strength (ACS) schedule that scales the drawing influence over denoising timesteps. Experiments on COCO and Sketchy report quantitative comparisons against SD, ControlNet, T2I-Adapter, UniControl, GLIGEN, and InstanceDiffusion, plus a user study indicating preference for VersaGen.
Significance. If the reported capabilities hold, VersaGen would be a practical step toward flexible, user-centric control in text-to-image generation, addressing a real limitation of scene-level control methods: users often want to control only part of the scene. The paper provides a public code link, a human study, and ablations of the main components, and the overall idea of partial/drawing-level control with automatic localization is timely and of broad interest to the controllable-generation community. The central mechanism is plausible and the empirical comparisons, while needing some scrutiny, do provide independent grounding. However, several technical gaps currently prevent full assessment of the method as described.
major comments (4)
- [Sec. 4.2 (Eq. 2)] The inference-time correspondence between each user drawing s_i and a text token c_i is assumed rather than specified. The paper states 'Assuming the textual tokens correspond to s_i are c_i' but offers no mechanism by which a user or the system determines which token names each drawing when a prompt and multiple drawings are supplied. The training-time procedure using spaCy and GroundingDINO (Sec. A.1) derives token indices from captions and ground-truth detections, which is not available at inference. Without this mapping, the localization map in Eq. 2, the OTSU mask R_i, the bounding box B_i, and the MCR/MoD objectives in Eqs. 3, 7, 9, 11-12 cannot be executed for arbitrary multi-subject or combination inputs. The Table 1 results for the 'without GT' rows could therefore depend on an oracle assignment, and the central claim of versatile multi-subject control is not yet supported for unconstrained user input. Please specify a concrete inference-time assignment procedure (e.g., vision-language matching between each drawing and the prompt's object nouns, or explicit user selection) and report results under that protocol.
- [Sec. 4.2 (Eq. 2)] The definition of the cross-attention map A_t^i(s_i, c_i) uses Q_n(s_i), a query vector derived from the drawing latent, but the cross-attention in the frozen U-Net is computed from the noisy latent z_t and the text keys. Later, Eq. 7 uses A_t^i(z_t, c_i) without explaining how this relates to Eq. 2. As written, the localization mechanism is under-specified and not reproducible: the paper does not state whether the drawing latent is concatenated with z_t (as Eq. 5 suggests), whether the attention map is taken from the trainable copy or the frozen U-Net, or how the query for the drawing is computed in the same embedding space. Please clarify the exact computation of the attention map, including the roles of s_i, z_t, and the text token c_i.
- [Sec. 4.4 (Eq. 13)] The adaptive control strength schedule γ(t) = 1 - a/(1+e^{-b(t-c)}) is monotonically increasing in t. Since the denoising process goes from t=T (noise) to t=0 (clean), this schedule makes the drawing influence increase as denoising progresses, with γ(0) ≈ 1.0 and γ(T) ≈ 0.3 for the given hyperparameters. This directly contradicts the text, which states that 'the influence of user drawings' is reduced as denoising progresses and that the textual prompt takes over in the later detail-generation stage. Either the formula is meant to be γ(t) = a/(1+e^{-b(t-c)}), which would decrease with t, or the description needs to be corrected. This is load-bearing because ACS is one of the three core claimed contributions and its ablation in Sec. 5.4 presumably tests the intended (decreasing) behavior.
- [Sec. 5.2 (Table 1)] The quantitative comparison with ControlNet, T2I-Adapter, and UniControl uses the automatically identified bounding boxes B'' or relocated drawings S' from VersaGen as the visual conditions for these baselines. These models are designed for and trained on holistic scene-level conditions (full edge maps or full scene control); presenting sparse object-level crops is out-of-distribution and likely depresses their performance. Conversely, VersaGen uses its own optimal partial-input protocol, making the comparison favorable to the proposed method. To support the claim that VersaGen 'outperforms' existing controllable T2I models, the paper should also report the baselines under their native full-scene condition (e.g., full edge map of the ground-truth image), and discuss the trade-off between information amount and capability. Without this, the headline quantitative superiority is not fully established.
minor comments (6)
- [Sec. 3 (Fig. 2)] The pilot study uses FID as an 'Objective Similarity' metric to compare user-submitted results with reference images. FID is a distribution-level metric that is not meaningful for single-image comparisons; please clarify whether the score is computed over the entire set of trials (which would be acceptable) and specify the number of images used. The reported values (0.173 and 0.217) are far below typical FID ranges, suggesting a different scale or computation; please document the exact protocol.
- [Sec. 4.2 (Eq. 2)] The dimensions of Q_n(s_i) and K_n(c_i) are written as R^{B*HW*d} and R^{B*d*77}. The 77-length dimension corresponds to the full text sequence in SD, but the key for a single token c_i should be a vector rather than a sequence. Please clarify whether the attention map is averaged over all text positions or computed only for the specific token index, and define the indexing convention.
- [Sec. 4.3 (Eq. 7)] The symbol L is used both for the number of layers in the summation (l=1..L) and for the loss (L_token, L_pixel, L). Using the same symbol for both is confusing; please rename one of them (e.g., N_l for the number of layers).
- [Sec. 2] The related work mentions 'Imegen' (first paragraph); the correct model name is 'Imagen'. Please fix the typo.
- [Sec. A.3 (Table 5)] The description of adapting VersaGen to SDXL is ambiguous. It states that 'only two MCR hyperparameters' need tuning and that the same hyperparameter configuration from the SD2.1 experiments was used, but it is unclear whether the User Drawing Encoder and fusion network were retrained on SDXL or whether the SD2.1-trained weights were used directly. Since the latent spaces differ, please clarify the training/adaptation procedure.
- [Sec. 5.4 (Fig. 6)] The ablation study is only qualitative. Adding a quantitative ablation table with the same metrics as Table 1 (e.g., with each of MCR, MoD, ACS removed) would strengthen the claim that each strategy contributes meaningfully.
Circularity Check
No significant circularity: VersaGen's central claims are empirically benchmarked against external baselines, and the only self-references are pointers to the paper's own supplementary material rather than load-bearing derivations.
full rationale
VersaGen is an empirical system paper rather than a formal derivation. The main claim—that four levels of visual control (single subject, multiple subjects, background, and combinations) improve controllable T2I generation—is validated by quantitative comparisons against external methods (Stable Diffusion, T2I-Adapter, ControlNet, UniControl, GLIGEN, InstanceDiffusion) on COCO and Sketchy (Table 1), with metrics such as CLIP, FID, DINO, and ACC computed against ground-truth images and external detectors. The core architecture (Sec. 4.1–4.2) adapts the externally published ControlNet pipeline onto a frozen Stable Diffusion backbone, and the inference-time strategies (MCR, MoD, ACS) are defined by explicit equations (Eqs. 7–13) whose inputs are user drawings, prompts, and cross-attention maps; none of these equations is defined in terms of the headline results. The hyperparameters (a, b, c, alpha, beta, lambda, tau) are set by hand rather than fitted to reproduce the reported metrics, so they are not predictions of those metrics. The only self-references (Chen et al. 2024) point to the paper's own supplementary material for implementation details (Sec. 4.4 and Sec. 5) and are not used to justify the main capability claims, which are independently grounded. The passage 'Assuming the textual tokens correspond to s_i are c_i' (Sec. 4.2, Eq. 2) is indeed a missing specification of how the system maps each user drawing to a text token at inference, but it is an incompleteness in the method description rather than a circular reduction: it does not make any output equal to an input by construction, and it does not borrow an unverified result from the authors' prior work. No step in the claimed derivation chain is equivalent to its own input or to a fitted parameter renamed as a prediction. Accordingly, no circularity step is identified.
Assumptions & free parameters
free parameters (7)
- acs_schedule_a =
0.7
- acs_schedule_b =
0.6
- acs_schedule_c =
15
- mcr_step_size_alpha =
15
- mcr_loss_lambda =
0.05
- mod_iou_threshold_beta =
0.1
- denoising_timestep_tau =
48
assumptions (6)
- domain assumption Cross-attention maps in Stable Diffusion localize object tokens and can be thresholded with OTSU to yield object masks.
- domain assumption Edge maps extracted via SAM segmentation, GroundingDINO detection, and Canny-style pseudo-graffiti are sufficiently representative of human drawings for training.
- domain assumption The correspondence between each input drawing s_i and its text token c_i is available at inference.
- domain assumption Stable Diffusion V2.1 is a reliable frozen foundation, and its attention maps provide enough signal for drawing placement.
- domain assumption Grounding DINO detections (box threshold 0.3) and DINOv2 features provide valid evaluation of object presence and perceptual similarity.
- domain assumption FID computed on the full test set is a meaningful measure of generation quality for this conditional generation task.
Cite this review
Pith. "Pith review of VersaGen: Unleashing Versatile Visual Control for Text-to-Image Synthesis." pith.science (2026). https://pith.science/paper/CLKOT64M
@misc{pith2026241211594,
author = {Pith},
title = {Pith review of: VersaGen: Unleashing Versatile Visual Control for Text-to-Image Synthesis},
year = {2026},
howpublished = {\url{https://pith.science/paper/CLKOT64M}},
note = {Machine review of arXiv:2412.11594}
}
read the original abstract
Despite the rapid advancements in text-to-image (T2I) synthesis, enabling precise visual control remains a significant challenge. Existing works attempted to incorporate multi-facet controls (text and sketch), aiming to enhance the creative control over generated images. However, our pilot study reveals that the expressive power of humans far surpasses the capabilities of current methods. Users desire a more versatile approach that can accommodate their diverse creative intents, ranging from controlling individual subjects to manipulating the entire scene composition. We present VersaGen, a generative AI agent that enables versatile visual control in T2I synthesis. VersaGen admits four types of visual controls: i) single visual subject; ii) multiple visual subjects; iii) scene background; iv) any combination of the three above or merely no control at all. We train an adaptor upon a frozen T2I model to accommodate the visual information into the text-dominated diffusion process. We introduce three optimization strategies during the inference phase of VersaGen to improve generation results and enhance user experience. Comprehensive experiments on COCO and Sketchy validate the effectiveness and flexibility of VersaGen, as evidenced by both qualitative and quantitative results.
Figures
Figures from the paper (6 more)
Reference graph
Works this paper leans on
-
[1]
Aubert, M.; Brumm, A.; Ramli, M.; Sutikna, T.; Saptomo, E. W.; Hakim, B.; Morwood, M. J.; van den Bergh, G. D.; Kinsley, L.; and Dosseto, A. 2014. Pleistocene cave art from Sulawesi, Indonesia. In Nature
work page 2014
-
[2]
Balaji, Y.; Nah, S.; Huang, X.; Vahdat, A.; Song, J.; Zhang, Q.; Kreis, K.; Aittala, M.; Aila, T.; Laine, S.; et al. 2022. ediff-i: Text-to-image diffusion models with an ensemble of expert denoisers. In arXiv preprint arXiv:2211.01324
arXiv 2022
-
[3]
Cao, M.; Wang, X.; Qi, Z.; Shan, Y.; Qie, X.; and Zheng, Y. 2023. Masactrl: Tuning-free mutual self-attention control for consistent image synthesis and editing. In ICCV
work page 2023
-
[4]
Chang, H.; Zhang, H.; Barber, J.; Maschinot, A.; Lezama, J.; Jiang, L.; Yang, M.-H.; Murphy, K.; Freeman, W. T.; Rubinstein, M.; et al. 2023. Muse: Text-to-image generation via masked generative transformers. In arXiv preprint arXiv:2301.00704
arXiv 2023
-
[5]
Chefer, H.; Alaluf, Y.; Vinker, Y.; Wolf, L.; and Cohen-Or, D. 2023. Attend-and-excite: Attention-based semantic guidance for text-to-image diffusion models. In ACM TOG
work page 2023
-
[6]
Chen, M.; Laina, I.; and Vedaldi, A. 2024. Training-free layout control with cross-attention guidance. In WACV
work page 2024
-
[7]
Chen, Z.; Yang, L.; Qi, Y.; Zhang, H.; Pang, K.; Li, K.; and Song, Y.-Z. 2024. VersaGen: Unleashing Versatile Visual Control for Text-to-Image Synthesis. In arXiv preprint arXiv:2412.11594
work page Pith review arXiv 2024
-
[8]
Eitz, M.; Hays, J.; and Alexa, M. 2012. How do humans sketch objects? In ACM TOG
work page 2012
Show all 55 references
-
[9]
E.; and Wang, W
Feng, W.; He, X.; Fu, T.-J.; Jampani, V.; Akula, A.; Narayana, P.; Basu, S.; Wang, X. E.; and Wang, W. Y. 2022. Training-free structured diffusion guidance for compositional text-to-image synthesis. In arXiv preprint arXiv:2212.05032
2022 arXiv
-
[10]
Gafni, O.; Polyak, A.; Ashual, O.; Sheynin, S.; Parikh, D.; and Taigman, Y. 2022. Make-a-scene: Scene-based text-to-image generation with human priors. In ECCV
2022
-
[11]
H.; Chechik, G.; and Cohen-Or, D
Gal, R.; Alaluf, Y.; Atzmon, Y.; Patashnik, O.; Bermano, A. H.; Chechik, G.; and Cohen-Or, D. 2022. An image is worth one word: Personalizing text-to-image generation using textual inversion. In arXiv preprint arXiv:2208.01618
2022 arXiv
-
[12]
F.; Naseer, M.; Khan, S.; and Wonka, P
Gani, H.; Bhat, S. F.; Naseer, M.; Khan, S.; and Wonka, P. 2023. Llm blueprint: Enabling text-to-image generation with complex and detailed prompts. In arXiv preprint arXiv:2310.10640
2023 arXiv
-
[13]
Ge, S.; Park, T.; Zhu, J.-Y.; and Huang, J.-B. 2023. Expressive text-to-image generation with rich text. In ICCV
2023
-
[14]
H.; and Gombrich, E
Gombrich, E. H.; and Gombrich, E. 1995. The story of art. Phaidon Press
1995
-
[15]
Gong, B.; Huang, S.; Feng, Y.; Zhang, S.; Li, Y.; and Liu, Y. 2023. Check, Locate, Rectify: A Training-Free Layout Calibration System for Text-to-Image Generation. In arXiv preprint arXiv:2311.15773
2023 arXiv
-
[16]
Hertz, A.; Mokady, R.; Tenenbaum, J.; Aberman, K.; Pritch, Y.; and Cohen-Or, D. 2023. Prompt-to-prompt image editing with cross attention control. In ICLR
2023
-
[17]
Heusel, M.; Ramsauer, H.; Unterthiner, T.; Nessler, B.; and Hochreiter, S. 2017. Gans trained by a two time-scale update rule converge to a local nash equilibrium. In NeurIPS
2017
-
[18]
L.; Standish, C
Hoffmann, D. L.; Standish, C. D.; Garc \' a-Diez, M.; Pettitt, P. B.; Milton, J. A.; Zilh \ a o, J.; Alcolea-Gonz \'a lez, J. J.; Cantalejo-Duarte, P.; Collado, H.; De Balb \' n, R.; et al. 2018. U-Th dating of carbonate crusts reveals Neandertal origin of Iberian cave art. In Science
2018
-
[19]
Honnibal, M.; Montani, I.; Van Landeghem, S.; and Boyd, A. 2020. spaCy: Industrial-strength Natural Language Processing in Python
2020
-
[20]
Jongejan, J.; Rowley, H.; Kawashima, T.; Kim, J.; and Fox-Gieg, N. 2017. The quick, draw! dataset
2017
-
[21]
Ju, X.; Zeng, A.; Zhao, C.; Wang, J.; Zhang, L.; and Xu, Q. 2023. HumanSD: A native skeleton-guided diffusion model for human image generation. In ICCV
2023
-
[22]
Karmiloff-Smith, A. 1990. Constraints on representational change: Evidence from children's drawing. In Cognition
1990
-
[23]
Kim, G.; Kwon, T.; and Ye, J. C. 2022. Diffusionclip: Text-guided diffusion models for robust image manipulation. In CVPR
2022
-
[24]
C.; Lo, W.-Y.; et al
Kirillov, A.; Mintun, E.; Ravi, N.; Mao, H.; Rolland, C.; Gustafson, L.; Xiao, T.; Whitehead, S.; Berg, A. C.; Lo, W.-Y.; et al. 2023. Segment anything. In ICCV
2023
-
[25]
Lee, K.; Liu, H.; Ryu, M.; Watkins, O.; Du, Y.; Boutilier, C.; Abbeel, P.; Ghavamzadeh, M.; and Gu, S. S. 2023. Aligning text-to-image models using human feedback. In arXiv preprint arXiv:2302.12192
2023 arXiv
-
[26]
Li, J.; Li, D.; Xiong, C.; and Hoi, S. 2022. Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation. In ICML
2022
-
[27]
Li, Y.; Keuper, M.; Zhang, D.; and Khoreva, A. 2023 a . Divide & bind your attention for improved generative semantic nursing. In arXiv preprint arXiv:2307.10864
2023 arXiv
-
[28]
Li, Y.; Liu, H.; Wu, Q.; Mu, F.; Yang, J.; Gao, J.; Li, C.; and Lee, Y. J. 2023 b . Gligen: Open-set grounded text-to-image generation. In CVPR
2023
-
[29]
Lin, T.-Y.; Maire, M.; Belongie, S.; Hays, J.; Perona, P.; Ramanan, D.; Doll \'a r, P.; and Zitnick, C. L. 2014. Microsoft coco: Common objects in context. In ECCV
2014
-
[30]
Liu, S.; Zeng, Z.; Ren, T.; Li, F.; Zhang, H.; Yang, J.; Li, C.; Yang, J.; Su, H.; Zhu, J.; et al. 2023. Grounding dino: Marrying dino with grounded pre-training for open-set object detection. In arXiv preprint arXiv:2303.05499
2023 arXiv
-
[31]
Liu, V.; and Chilton, L. B. 2022. Design guidelines for prompt engineering text-to-image generative models. In CHI
2022
-
[32]
Loshchilov, I.; and Hutter, F. 2019. Decoupled Weight Decay Regularization. In ICLR
2019
-
[33]
Ma, J.; Liang, J.; Chen, C.; and Lu, H. 2023. Subject-diffusion: Open domain personalized text-to-image generation without test-time fine-tuning. In arXiv preprint arXiv:2307.11410
2023 arXiv
-
[34]
Marcus, G.; Davis, E.; and Aaronson, S. 2022. A very preliminary analysis of DALL-E 2. In arXiv preprint arXiv:2204.13807
2022 arXiv
-
[35]
Mou, C.; Wang, X.; Xie, L.; Wu, Y.; Zhang, J.; Qi, Z.; and Shan, Y. 2024. T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. In AAAI
2024
-
[36]
Oquab, M.; Darcet, T.; Moutakanni, T.; Vo, H.; Szafraniec, M.; Khalidov, V.; Fernandez, P.; Haziza, D.; Massa, F.; El-Nouby, A.; et al. 2023. Dinov2: Learning robust visual features without supervision. In arXiv preprint arXiv:2304.07193
2023 arXiv
-
[37]
Otsu, N.; et al. 1975. A threshold selection method from gray-level histograms. In Automatica
1975
-
[38]
Patashnik, O.; Garibi, D.; Azuri, I.; Averbuch-Elor, H.; and Cohen-Or, D. 2023. Localizing object-level shape variations with text-to-image diffusion models. In ICCV
2023
-
[39]
Podell, D.; English, Z.; Lacey, K.; Blattmann, A.; Dockhorn, T.; M \"u ller, J.; Penna, J.; and Rombach, R. 2023. SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis. In ICLR
2023
-
[40]
C.; Xiong, C.; Savarese, S.; et al
Qin, C.; Zhang, S.; Yu, N.; Feng, Y.; Yang, X.; Zhou, Y.; Wang, H.; Niebles, J. C.; Xiong, C.; Savarese, S.; et al. 2024. UniControl: A Unified Diffusion Model for Controllable Visual Generation In the Wild. In NeurIPS
2024
-
[41]
W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al
Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In ICML
2021
-
[42]
Ramesh, A.; Pavlov, M.; Goh, G.; Gray, S.; Voss, C.; Radford, A.; Chen, M.; and Sutskever, I. 2021. Zero-shot text-to-image generation. In ICML
2021
-
[43]
Rassin, R.; Hirsch, E.; Glickman, D.; Ravfogel, S.; Goldberg, Y.; and Chechik, G. 2024. Linguistic binding in diffusion models: Enhancing attribute correspondence through attention map alignment. In NeurIPS
2024
-
[44]
Rombach, R.; Blattmann, A.; Lorenz, D.; Esser, P.; and Ommer, B. 2022. High-resolution image synthesis with latent diffusion models. In CVPR
2022
-
[45]
L.; Ghasemipour, K.; Gontijo Lopes, R.; Karagol Ayan, B.; Salimans, T.; et al
Saharia, C.; Chan, W.; Saxena, S.; Li, L.; Whang, J.; Denton, E. L.; Ghasemipour, K.; Gontijo Lopes, R.; Karagol Ayan, B.; Salimans, T.; et al. 2022. Photorealistic text-to-image diffusion models with deep language understanding. In NeurIPS
2022
-
[46]
Sangkloy, P.; Burnell, N.; Ham, C.; and Hays, J. 2016. The sketchy database: learning to retrieve badly drawn bunnies. In ACM TOG
2016
-
[47]
S.; Girdhar, R.; and Misra, I
Wang, X.; Darrell, T.; Rambhatla, S. S.; Girdhar, R.; and Misra, I. 2024. Instancediffusion: Instance-level control for image generation. In CVPR, 6232--6242
2024
-
[48]
Wang, Z.; Sha, Z.; Ding, Z.; Wang, Y.; and Tu, Z. 2023. TokenCompose: Grounding Diffusion with Token-level Supervision. In arXiv preprint arXiv:2312.03626
2023 arXiv
-
[49]
J.; Montoya, E.; Munechika, D.; Yang, H.; Hoover, B.; and Chau, D
Wang, Z. J.; Montoya, E.; Munechika, D.; Yang, H.; Hoover, B.; and Chau, D. H. 2022. Diffusiondb: A large-scale prompt gallery dataset for text-to-image generative models. In arXiv preprint arXiv:2210.14896
2022 arXiv
-
[50]
Wu, Q.; Liu, Y.; Zhao, H.; Bui, T.; Lin, Z.; Zhang, Y.; and Chang, S. 2023. Harnessing the spatial-temporal attention of diffusion models for high-fidelity text-to-image synthesis. In ICCV
2023
-
[51]
Xu, J.; Liu, X.; Wu, Y.; Tong, Y.; Li, Q.; Ding, M.; Tang, J.; and Dong, Y. 2024. Imagereward: Learning and evaluating human preferences for text-to-image generation. In NeurIPS
2024
-
[52]
Zhang, K.; Li, Y.; Zuo, W.; Zhang, L.; Van Gool, L.; and Timofte, R. 2021. Plug-and-play image restoration with deep denoiser prior. In TPAMI
2021
-
[53]
Zhang, L.; Rao, A.; and Agrawala, M. 2023. Adding conditional control to text-to-image diffusion models. In ICCV
2023
-
[54]
, " * write output.state after.block = add.period write newline
ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...
-
[55]
write newline
" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...
Reviewed August 11, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.