Pith. sign in

REVIEW 5 major objections 6 minor 74 references

From Missing Pieces to Masterpieces: Image Completion with Context-Adaptive Diffusion

T0 review · 5 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read ConFill claims a context-adaptive discrepancy term, inserted into a fixed pre-trained diffusion model, produces more coherent image completions than existing inpainting methods.

desk verdict A plausible new discrepancy term with substantial experiments, but the posterior derivation is circular enough that the headline gains are not yet traceable to the mechanism. read the letter →

arxiv 2504.14294 v1 pith:MA5UW5MU submitted 2025-04-19 cs.CV

Pourya Shamsolmoali , Masoumeh Zareapoor , Huiyu Zhou , Michael Felsberg , Dacheng Tao , Xuelong Li This is my paper · ORCID
classification cs.CV
keywords imagecompletioninpaintingdiffusionmodelscontext-adaptivediscrepancyoptimaltransportdynamicsamplinggenerativeinverseproblemsunsupervised
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

ConFill is an image-completion framework that leaves the diffusion model itself untouched and instead changes how the denoising loop measures and enforces agreement with the known pixels. Its central claim is that existing diffusion inpainting methods fail because their discrepancy metrics, Euclidean distance or plain Wasserstein distance, ignore local context, so the evolving distribution of the missing region never aligns with the surrounding image. The paper replaces those metrics with a Context-Adaptive Discrepancy (CAD) that scales an optimal-transport cost by local edge, texture, and semantic differences, and adds a Dynamic Sampling rule that allocates more denoising steps to complex regions. If the claim is right, a pre-trained diffusion model becomes the best tested unsupervised image completer: on ImageNet-1K the reported LPIPS is 10.9% lower than RePaint and 6.7% lower than CoPaint, with better FID on all three datasets.

What carries the argument

The load-bearing object is the Context-Adaptive Discrepancy, $W_C(x,y;C)=\|\nabla\varphi(x)-y\|^2 \cdot f(C(x),C(y))$, where $\nabla\varphi$ is the gradient of a Brenier potential giving the optimal transport map between latent distributions, and $f(C(x),C(y))=1+\upsilon e^{-\tau\|C(x)-C(y)\|}$ is a scaling factor that up-weights the cost when local contextual features differ. The contextual features $C(\cdot)$ combine handcrafted edge and texture filters with a pre-trained semantic feature extractor. This term does two jobs: it replaces the Euclidean discrepancy in the diffusion loss, and it defines the completion constraint in the approximate posterior of Eq. (13). The supporting mechanism is Dynamic Sampling, whose adaptability function $\varrho(x)=\hat{\alpha}\exp(-\psi\,\mathrm{Var}(x)N/(N-1))+\hat{\beta}\,\mathrm{ED}(x)$ sets the number of samples per region, concentrating computational effort where textural and structural complexity is highest.

What would settle it

A direct check is to replace Eq. (14)'s variance with a constant and re-run the ImageNet experiment; if the LPIPS gain over RePaint shrinks by less than half, the adaptive calibration is not responsible for the reported improvement.

Watch

Extended reading notes

Core claim

ConFill's core discovery is that the completion constraint can be pushed into the very first latent $\hat{x}_T$ instead of being enforced by replacement or blending at every step. The posterior $p_\theta(\hat{x}_T \mid O)$ is written as the prior times a CAD-aware delta function (Eq. 10), and its log-density is approximated by a Gaussian whose variance $\gamma'^2_T$ is estimated from the CAD values of the target image (Eqs. 13-14). During denoising, each step minimizes a loss that combines the standard transition term with $\gamma_t \cdot W_C(s(g_\theta(x_t)), r_0; C)$, where $W_C$ is the context-scaled optimal-transport discrepancy of Eqs. (6)-(7). A Dynamic Sampling schedule then raises the sample density in high-variance, high-edge-density regions. Taken together, the method progressively reduces the discrepancy between known and unknown regions at every timestep, which the paper argues is what produces the reported gains in fidelity and coherence.

Load-bearing premise

The method assumes that one number, a variance estimated from how well the model currently matches the known pixels, can correctly calibrate the completion constraint for every part of the image at every denoising step; if this single calibration is wrong, the guidance is miscalibrated and the reported gains do not follow.

Editorial extensions

If this is right

  • With ConFill, a pre-trained diffusion model can be converted into a strong image completer without retraining on masks; the paper reports best average FID, LPIPS, and SSIM over all tested baselines on CelebA-HQ, Places2, and ImageNet-1K.
  • The method generalizes to mask types never seen during tuning: on image-expansion and nearest-neighbour masks, ConFill achieves FID 19.63 versus 23.87 for CoPaint, a 17.8% improvement.
  • Dynamic sampling makes the framework faster than the resampling baselines: about 109 seconds per image on unseen masks versus roughly 211 for RePaint and 453 for DPS.
  • Human preference tracks the metric gains: ConFill was 21% more likely to be selected over CoPaint on Places2 at mask ratio 0.8-1.0 and 29% more likely over RePaint on ImageNet at mask ratio 0.4-0.6.

Reading between the lines

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

  • An implication the paper leaves implicit is that the variance calibration in Eq. (14) is estimated from the same CAD values the method is trying to minimize; a natural test is to replace $\gamma'^2_T$ with a fixed constant and measure whether the LPIPS gains survive, which would isolate whether the adaptive scaling or the CAD term itself is doing the work.
  • Because the CAD formulation only assumes a differentiable generator and a known-region operator $s(\cdot)$, the same guidance could in principle be applied to other inverse problems such as super-resolution or deblurring; that extension is not tested in the paper.
  • The per-mask tables show the largest gains on the hardest masks (Expand, Half); on Narrow masks the gap to RePaint and CoPaint shrinks considerably, suggesting ConFill's advantage is concentrated where the completion problem is underdetermined.
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. The paper proposes ConFill, an unsupervised image completion framework built on a pre-trained diffusion model. The method introduces a Context-Adaptive Discrepancy (CAD) model that integrates local context into an optimal-transport-based discrepancy measure, which is used to guide the reverse diffusion process toward satisfying the completion constraint. A dynamic sampling scheme is also proposed to increase sampling density in regions with high textural or structural complexity. The authors report state-of-the-art performance on CelebA-HQ, Places2, and ImageNet-1K across several mask types, including LPIPS improvements of 10.9% over RePaint and 6.7% over CoPaint, better FID on all three datasets, and higher human preference in a user study.

Significance. If the method worked as described, it would be a meaningful advance in training-free image completion: a context-adaptive discrepancy metric that replaces hand-tuned guidance in diffusion-based inversion, with empirical gains over strong baselines. The experimental section is broad, covering many baselines, mask types, ablations, and a human study, and the authors report multiple metrics. However, the theoretical derivation of the CAD-guided posterior is the core of the contribution, and it contains serious formal flaws that undermine the link between the CAD mechanism and the reported results. The paper does not release code, and the described algorithm is not precisely specified, so the empirical claims cannot currently be traced to the proposed mechanism. As presented, the central claim is therefore not established.

major comments (5)
  1. [Sec. 3.2, Eq. (14)] The variance γ'^2_T is calibrated using the same CAD values and the same target r0 that the completion constraint is meant to enforce. At inference time the full target is unavailable, and if the calibration uses the masked known region alone, a single global scalar cannot adapt per region and timestep as claimed. Moreover, a likelihood variance should be a mean squared residual, not a mean CAD value; using mean CAD as γ'^2 mis-scales the guidance. The sampling loop is therefore not a valid posterior sampler under the stated model, and the reported gains cannot be traced to CAD.
  2. [Sec. 3.4, Eq. (19)] The second term applies CAD to a single argument, 'CAD( r0 − s(g_theta^(t−1)(hat_x_{t−1})) )^2', but CAD is defined in Eq. (16) as a sum over pointwise discrepancies between two distributions. The difference of two images is not a distribution, so this expression is undefined. The derivation of Eq. (19) is therefore incomplete and cannot support the claimed gradual approximation.
  3. [Algorithm 2, lines 6 and 9] The update hat_x_{t−1} = g_theta^(t)(hat_x_t) − λ∇φ_t(hat_x_t) is not the gradient step on L_CAD as defined in Eq. (9), which would be hat_x_{t−1} = hat_x_t − λ∇_{hat_x_t} L_CAD. Additionally, the loss described in line 6 (minimizing 'CAD(hat_x_t, r0) + λ_t ||∇φ_t(hat_x_t) − hat_x_t||') does not match Eq. (9). This mismatch makes it impossible to identify which objective is actually optimized in the experiments, so the connection between the posterior derivation and the reported performance is broken.
  4. [Sec. 3.2, Eqs. (10)–(13)] The replacement of the observation likelihood with a delta function and the subsequent Gaussian approximation are asserted without a limiting argument. The statement following Eq. (11) that the approximation becomes increasingly accurate as γ_T → 0 is in tension with Eq. (14), which sets γ' to be 'sufficiently large'. This inconsistency needs to be resolved for the posterior derivation to be coherent.
  5. [Sec. 4.1 and Eq. (14)] The hyperparameters are selected on the first five validation images, and Eq. (14) is evaluated per-sample using r0. It is unclear whether r0 in Eq. (14) is the full ground-truth image or the observed (masked) region only. If the former, the method leaks test-time information; if the latter, the global scalar γ' cannot provide the per-region adaptation claimed. Please clarify the exact input used in the calibration.
minor comments (6)
  1. [Throughout] There are numerous typos and inconsistent notations, including 'PowePaint' (Sec. 4.4), 'COPaint' (Sec. 4.4), 'p′θ(ˆxT|OT)' (Sec. 3.4), and the overloaded use of N in Eqs. (14) and (15) (as dimension, sample count, and pixel count).
  2. [Sec. 3.2, Eq. (4)] The definition of L_diffusion is not used elsewhere in the paper; consider removing it or explicitly connecting it to the CAD loss.
  3. [Algorithm 1] The transport map f computed in Algorithm 1 is not referenced in the update equations of Algorithm 2, so its role in the overall method is unclear.
  4. [Fig. 2] The figure is not referenced in the text, and the curves are not clearly labeled; the caption says 'mean pixel-wise discrepancy' but it is unclear whether this is computed for CAD, Wasserstein, or Euclidean measures, and over which images.
  5. [Sec. 4.6, Table 4] The paper states that J=1 and ´t=10 are used in the experiments, but Table 4 does not show a row for J=1 and ´t=15 or other combinations; please report the full grid used for selection.
  6. [Abstract] The phrase 'setting a new benchmark' overstates the contribution of a single paper; a more cautious wording, such as 'state-of-the-art results on the evaluated datasets', would be more appropriate.

Circularity Check

1 steps flagged · score 4.0 of 10

Eq. (14) makes the posterior variance a function of the very CAD residual it scales, so the CAD-guidance derivation is partially self-referential; the empirical benchmark claims remain externally evaluated.

  1. self definitional [Sec. 3.2, Eq. (14), used in Eq. (13)]
    "γ′2 T = 1 N PN i=1CAD s g(T) θ (ˆx(i) T ) , s (ˆx(i) 0 ) , (14) ... we adjust γ′2 T based on the average CAD across samples, where N represents the r0’s dimension. Here, CAD(·,·) measures the shortest distance between the transformed versions of ˆxT and ˆx0."

    In Eq. (13), the completion guidance uses precision 1/γ′2_T to scale the squared CAD between the known region r0 and the one-step estimate s(g_θ^T(ˆx_T)). Eq. (14) then defines γ′2_T as the average of the same CAD discrepancy between that one-step estimate and s(ˆx_0), i.e., the known region r0 that the completion constraint is enforcing. The variance is therefore not a fixed posterior dispersion: it is a function of the very residual it weights. A poor estimate enlarges γ′2_T and down-weights the constraint, while a good estimate shrinks it and amplifies the constraint. Consequently the 'accurate approximation of the posterior' in Eq.

full rationale

The CAD-guidance derivation has one self-referential step: the variance parameter γ′2_T in Eq. (13) is set in Eq. (14) from the same CAD discrepancy that the completion constraint measures, so the posterior approximation is an adaptive function of its own residual rather than a fixed Gaussian. This is load-bearing for the claimed derivation of CAD guidance. However, the paper's central empirical claim, state-of-the-art image completion, rests on external metrics (FID, LPIPS, SSIM, user study) and standard baselines, so the paper is not circular overall. The self-citations ([10], [12]) appear only in related-work context and do not carry the derivation; the Brenier and optimal-transport citations are external prior work. No other step reduces by construction. Score 4 reflects partial, not total, circularity.

Assumptions & free parameters 8 free parameters · 6 assumptions · 1 invented entities

The method rests on a context-weighted Brenier/Wasserstein discrepancy whose computation is not fully specified, an adaptive variance that is calibrated on the completion target itself, and multiple unstated hyperparameters (psi, m). No code or data release and no error bars supply external checks, so the reader must accept these premises on the authors' word.

free parameters (8)
  • upsilon = 0.1
    Context-scaling strength in Eq. (7); selected on the first five validation images and ablated in Table 6.
  • tau = 0.02
    Context-difference sensitivity in Eq. (7); selected on validation and ablated in Table 6.
  • psi = not reported
    Scaling parameter in the adaptability function Eq. (15); controls sensitivity to textural variation, no value is given.
  • alpha_hat, beta_hat = 0.5, 0.5 combined
    Weights for texture and edge complexity in Eq. (15); ablated in Table 5.
  • m = not reported
    Maximum samples per region in S(x)=ceil(m*rho(x)); no value is specified.
  • gamma_prime squared = computed from data via Eq. (14)
    Variance of the Gaussian posterior approximation is estimated from CAD values on the target image, not fixed independently.
  • learning rate lambda = 0.01 initial, adaptive
    Gradient step size for x_t updates; the adaptive schedule is not fully specified.
  • G, t_interval, J = G=2, t_interval=10, J=1
    Gradient descent steps, time-travel interval, and frequency; selected by the ablations in Tables 3-4.
assumptions (6)
  • domain assumption Brenier potential defines an optimal transport map between latent distributions at each diffusion timestep, computable in the image domain.
    Used in Eq. (6) and Algorithm 1; no proof is given that image patch distributions admit the required regularity or that the map is computable.
  • ad hoc to paper The posterior factorizes as p_theta(x_T) times a delta function enforcing s(x_0)=r_0 under CAD.
    Eq. (10) asserts this factorization; no normalization or limiting argument is provided.
  • ad hoc to paper A single adaptive variance gamma'^2_T computed from CAD values calibrates the completion constraint.
    Eq. (14) estimates the variance from the target r0 and the same CAD objective, which makes the approximation self-referential.
  • domain assumption Sobel, Gabor, and ResNet18 features capture the structural and semantic context needed for completion.
    The feature extractor C(.) in Eqs. (6)-(7) is used without quantifying feature quality beyond the handcrafted versus learned ablation in Table 7.
  • domain assumption Local variance and edge density are sufficient measures of reconstruction complexity for dynamic sampling.
    Eq. (15) and Algorithm 1 assume these statistics predict where the diffusion process needs denser sampling.
  • domain assumption A pre-trained DDPM from RePaint [17] and ADM [66] is an accurate unconditional prior for all three datasets.
    Section 4.1 uses these models without fine-tuning for the base diffusion; errors in the prior propagate to every CAD-guided completion.
invented entities (1)
  • delta_CAD, the CAD-based delta function
    purpose: Formal device to enforce the exact completion constraint s(x_0)=r_0 in the posterior, Eqs. (10)-(13).
    No falsifiable prediction outside the paper; it is an unmeasured formal construct introduced to make the posterior approximation look exact.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Missing Pieces to Masterpieces: Image Completion with Context-Adaptive Diffusion." pith.science (2026). https://pith.science/paper/MA5UW5MU

@misc{pith2026250414294,
  author       = {Pith},
  title        = {Pith review of: From Missing Pieces to Masterpieces: Image Completion with Context-Adaptive Diffusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MA5UW5MU}},
  note         = {Machine review of arXiv:2504.14294}
}
read the original abstract

Image completion is a challenging task, particularly when ensuring that generated content seamlessly integrates with existing parts of an image. While recent diffusion models have shown promise, they often struggle with maintaining coherence between known and unknown (missing) regions. This issue arises from the lack of explicit spatial and semantic alignment during the diffusion process, resulting in content that does not smoothly integrate with the original image. Additionally, diffusion models typically rely on global learned distributions rather than localized features, leading to inconsistencies between the generated and existing image parts. In this work, we propose ConFill, a novel framework that introduces a Context-Adaptive Discrepancy (CAD) model to ensure that intermediate distributions of known and unknown regions are closely aligned throughout the diffusion process. By incorporating CAD, our model progressively reduces discrepancies between generated and original images at each diffusion step, leading to contextually aligned completion. Moreover, ConFill uses a new Dynamic Sampling mechanism that adaptively increases the sampling rate in regions with high reconstruction complexity. This approach enables precise adjustments, enhancing detail and integration in restored areas. Extensive experiments demonstrate that ConFill outperforms current methods, setting a new benchmark in image completion.

Figures

Figures reproduced from arXiv: 2504.14294 by the authors.

Figure 1
Figure 1. (a) Generated images by DPS, RePaint, and our ConFill, using a fixed diffusion model on different masked inputs. (b) LPIPS score vs. denoising timesteps for a single input image. denoising process [16], [17]. Similarly, [2] focuses on optimizing pixel distribution rather than denoising, aiming to reduce the number of iterations needed for training. Although these methods modify large portions of an image, they strug… view at source ↗
Figure 2
Figure 2. The evolution of the discrepancy during the denoising pro￾cess. This illustrates the mean pixel-wise discrepancy using Context￾Adaptive Discrepancy (CAD) (red), Wasserstein Discrepancy (blue), and Euclidean Distance (orange). probability distribution aligns with that of the DDPM framework. This denoising process can be approximated as a weighted-sum blending of the degraded image iteratively at each timestep. Even i… view at source ↗
Figure 3
Figure 3. The ConFill framework. The red curve represents the CAD inverse diffusion steps, which iteratively balance the distribution of image patches across the latent space. The green and red hills represent latent distributions closer to the final completed image and noisy initial steps, respectively. discrepancy metric dynamically adjusts to both local and global image features, providing a more accurate measure of simila… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison on the CelebA-HQ. The facial images generated by ConFill not only present more distinctive facial features but also show a higher degree of similarity and coherence with the original images compared to those produced by other baseline models. to …
Figure 7
Figure 7. Figure 7: Evaluation of FID (top row) and LPIPS (bottom row) scores for ConFill and baseline models on the CelebA-HQ and Places2 datasets. ConFill consistently outperforms the other models in both fidelity and perceptual similarity at all iteration levels. an adaptive learning r…
Figure 8
Figure 8. Figure 8: The results of the user study comparing our model, ConFill, with baseline models CoPaint, RePaint, and DPS across various mask ratios on three different datasets: CelebA, Places2, and ImageNet. ConFill consistently outperforms the other models across all datasets, as d…
Figure 9
Figure 9. Figure 9 [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 11
Figure 11. Figure 11: Ablation analysis using a half-mask on the CelebA and Places2 datasets. Shows the performance of ConFill with and without dynamic sampling (DS) and compares it with the other models. Results show ConFill with DS performs strongly even at lower NFEs. sampling compared …
Figure 12
Figure 12. Figure 12: Image completion results of ConFill for the CelebA and Places2 datasets using various mask types (small to large size). TABLE 3: Ablation Analysis. Performance evaluation of our model using CAD, standard WD and Euclidean distance with varying gradi￾ent descent steps (…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 72 canonical work pages

  1. [1]

    Editgan: High-precision semantic image editing,

    H. Ling, K. Kreis, D. Li, S. W. Kim, A. Torralba, and S. Fidler, “Editgan: High-precision semantic image editing,” Proc. Advances Neural Inf. Process. Syst., vol. 34, pp. 16 331–16 345, 2021

    work page 2021

  2. [2]

    Image Inpainting via Iteratively Decoupled Probabilistic Modeling

    W. Li, X. Yu, K. Zhou, Y . Song, Z. Lin, and J. Jia, “Image inpaint- ing via iteratively decoupled probabilistic modeling,” arXiv preprint arXiv:2212.02963, 2022. 13

    work page Pith review arXiv 2022

  3. [3]

    Adversarial cross-spectral face completion for nir-vis face recognition,

    R. He, J. Cao, L. Song, Z. Sun, and T. Tan, “Adversarial cross-spectral face completion for nir-vis face recognition,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 42, no. 5, pp. 1025–1037, 2020

    work page 2020

  4. [4]

    All-in-one image restoration for unknown corruption,

    B. Li, X. Liu, P. Hu, Z. Wu, J. Lv, and X. Peng, “All-in-one image restoration for unknown corruption,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2022, pp. 17 452–17 462

    work page 2022

  5. [5]

    Inst- inpaint: Instructing to remove objects with diffusion models,

    A. B. Yildirim, V . Baday, E. Erdem, A. Erdem, and A. Dundar, “Inst- inpaint: Instructing to remove objects with diffusion models,” arXiv preprint arXiv:2304.03246, 2023

    arXiv 2023

  6. [6]

    Resolution- robust large mask inpainting with fourier convolutions,

    R. Suvorov, E. Logacheva, A. Mashikhin, A. Remizova, A. Ashukha, A. Silvestrov, N. Kong, H. Goka, K. Park, and V . Lempitsky, “Resolution- robust large mask inpainting with fourier convolutions,” in Proc. Winter Conf. Appl. Comput. Vis., 2022, pp. 2149–2159

    work page 2022

  7. [7]

    Glama: Joint spatial and frequency loss for general image inpainting,

    Z. Lu, J. Jiang, J. Huang, G. Wu, and X. Liu, “Glama: Joint spatial and frequency loss for general image inpainting,” in Proc. Conf. Comput. Vis. Pattern Recognit. Workshop, 2022, pp. 1301–1310

    work page 2022

  8. [8]

    Aggregated contextual transfor- mations for high-resolution image inpainting,

    Y . Zeng, J. Fu, H. Chao, and B. Guo, “Aggregated contextual transfor- mations for high-resolution image inpainting,” IEEE Trans. Vis. Comput. Graph., 2022

    work page 2022

Show all 74 references
  1. [9]

    Mat: Mask-aware transformer for large hole image inpainting,

    W. Li, Z. Lin, K. Zhou, L. Qi, Y . Wang, and J. Jia, “Mat: Mask-aware transformer for large hole image inpainting,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2022, pp. 10 758–10 768

    2022

  2. [10]

    Transinpaint: Transformer-based image inpainting with context adaptation,

    P. Shamsolmoali, M. Zareapoor, and E. Granger, “Transinpaint: Transformer-based image inpainting with context adaptation,” in Proc. Int. Conf. Comput. Vis., 2023, pp. 849–858

    2023

  3. [11]

    High-fidelity pluralistic image completion with transformers,

    Z. Wan, J. Zhang, D. Chen, and J. Liao, “High-fidelity pluralistic image completion with transformers,” in Proc. Int. Conf. Comput. Vis., 2021

    2021

  4. [12]

    Vtae: Variational transformer autoencoder with manifolds learning,

    P. Shamsolmoali, M. Zareapoor, H. Zhou, D. Tao, and X. Li, “Vtae: Variational transformer autoencoder with manifolds learning,” IEEE Trans. Image Process., 2023

    2023

  5. [13]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Proc. Advances Neural Inf. Process. Syst., vol. 33, pp. 6840–6851, 2020

    2020

  6. [14]

    Denoising diffusion implicit models,

    J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” Proc. Int. Conf. Learn. Represent., 2021

    2021

  7. [15]

    Zero-shot image restoration using denoising diffusion null-space,

    Y . Wang, J. Yu, and J. Zhang, “Zero-shot image restoration using denoising diffusion null-space,” Proc. Int. Conf. Learn. Represent., 2023

    2023

  8. [16]

    Blended diffusion for text- driven editing of natural images,

    O. Avrahami, D. Lischinski, and O. Fried, “Blended diffusion for text- driven editing of natural images,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2022, pp. 18 208–18 218

    2022

  9. [17]

    Repaint: Inpainting using denoising diffusion probabilistic models,

    A. Lugmayr, M. Danelljan, A. Romero, F. Yu, R. Timofte, and L. Van Gool, “Repaint: Inpainting using denoising diffusion probabilistic models,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2022

    2022

  10. [18]

    Diffusion probabilistic modeling of protein backbones in 3d for the motif-scaffolding problem,

    B. L. Trippe, J. Yim, D. Tischer, D. Baker, T. Broderick, R. Barzilay, and T. Jaakkola, “Diffusion probabilistic modeling of protein backbones in 3d for the motif-scaffolding problem,” Proc. Advances Neural Inf. Process. Syst. workshop, 2022

    2022

  11. [19]

    On fast sampling of diffusion probabilistic models,

    Z. Kong and W. Ping, “On fast sampling of diffusion probabilistic models,” Proc. Int. Conf. Mach. Learn. workshop, 2021

    2021

  12. [20]

    Importance sampling cams for weakly- supervised segmentation,

    A. Jonnarth and M. Felsberg, “Importance sampling cams for weakly- supervised segmentation,” in Proc. Int. Conf. Acoust. Speech Signal Process., 2022, pp. 2639–2643

    2022

  13. [21]

    Pseudo numerical methods for diffusion models on manifolds,

    L. Liu, Y . Ren, Z. Lin, and Z. Zhao, “Pseudo numerical methods for diffusion models on manifolds,”Proc. Int. Conf. Learn. Represent., 2022

    2022

  14. [22]

    Diffusion earth mover’s distance and distribution embeddings,

    A. Y . Tong, G. Huguet, A. Natik, K. MacDonald, M. Kuchroo, R. Coif- man, G. Wolf, and S. Krishnaswamy, “Diffusion earth mover’s distance and distribution embeddings,” in Proc. Int. Conf. Mach. Learn., 2021

    2021

  15. [23]

    Dpm-ot: a new diffusion probabilistic model based on optimal transport,

    Z. Li, S. Li, Z. Wang, N. Lei, Z. Luo, and D. X. Gu, “Dpm-ot: a new diffusion probabilistic model based on optimal transport,” in Proc. Int. Conf. Comput. Vis., 2023, pp. 22 624–22 633

    2023

  16. [24]

    Wasserstein dependency measure for representation learning,

    S. Ozair, C. Lynch, Y . Bengio, A. Van den Oord, S. Levine, and P. Ser- manet, “Wasserstein dependency measure for representation learning,” Proc. Advances Neural Inf. Process. Syst., vol. 32, 2019

    2019

  17. [25]

    Diffusion posterior sampling for general noisy inverse problems,

    H. Chung, J. Kim, M. T. Mccann, M. L. Klasky, and J. C. Ye, “Diffusion posterior sampling for general noisy inverse problems,” Proc. Int. Conf. Learn. Represent., 2023

    2023

  18. [26]

    Improving diffusion models for inverse problems using manifold constraints,

    H. Chung, B. Sim, D. Ryu, and J. C. Ye, “Improving diffusion models for inverse problems using manifold constraints,” Proc. Advances Neural Inf. Process. Syst., vol. 35, pp. 25 683–25 696, 2022

    2022

  19. [27]

    Towards coherent image inpainting using denoising diffusion implicit models,

    G. Zhang, J. Ji, Y . Zhang, M. Yu, T. S. Jaakkola, and S. Chang, “Towards coherent image inpainting using denoising diffusion implicit models,” Proc. Int. Conf. Mach. Learn., 2023

    2023

  20. [28]

    Brenier approach for optimal transportation between a quasi-discrete measure and a discrete measure,

    Y . Lu, L. Chen, A. Saidi, and X. Gu, “Brenier approach for optimal transportation between a quasi-discrete measure and a discrete measure,” 2019, pp. 204–212

    2019

  21. [29]

    Regularity as regularization: Smooth and strongly convex brenier potentials in optimal transport,

    F.-P. Paty, A. d’Aspremont, and M. Cuturi, “Regularity as regularization: Smooth and strongly convex brenier potentials in optimal transport,” in Proc. Int. Conf. Artif. Intell. Statis., 2020, pp. 1222–1232

    2020

  22. [30]

    Optimal transport for image processing,

    N. Papadakis, “Optimal transport for image processing,” Ph.D. disserta- tion, Universit´e de Bordeaux; Habilitation thesis, 2015

    2015

  23. [31]

    Patch- match: A randomized correspondence algorithm for structural image editing,

    C. Barnes, E. Shechtman, A. Finkelstein, and D. B. Goldman, “Patch- match: A randomized correspondence algorithm for structural image editing,” ACM Trans. Graph., vol. 28, no. 3, p. 24, 2009

    2009

  24. [32]

    Globally and locally consis- tent image completion,

    S. Iizuka, E. Simo-Serra, and H. Ishikawa, “Globally and locally consis- tent image completion,” ACM Trans. Graph., vol. 36, no. 4, 2017

    2017

  25. [33]

    Generative image inpainting with contextual attention,

    J. Yu, Z. Lin, J. Yang, X. Shen, X. Lu, and T. S. Huang, “Generative image inpainting with contextual attention,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2018, pp. 5505–5514

    2018

  26. [34]

    Semantic image inpainting with progressive generative networks,

    H. Zhang, Z. Hu, C. Luo, W. Zuo, and M. Wang, “Semantic image inpainting with progressive generative networks,” in Proc. ACM Int. Conf. Multimedia, 2018, pp. 1939–1947

    2018

  27. [35]

    Rethinking image inpainting via a mutual encoder-decoder with feature equalizations,

    H. Liu, B. Jiang, Y . Song, W. Huang, and C. Yang, “Rethinking image inpainting via a mutual encoder-decoder with feature equalizations,” in Proc. Eur. Conf. Comput. Vis., 2020, pp. 725–741

    2020

  28. [36]

    Cr-fill: Generative image inpainting with auxiliary contextual reconstruction,

    Y . Zeng, Z. Lin, H. Lu, and V . M. Patel, “Cr-fill: Generative image inpainting with auxiliary contextual reconstruction,” in Proc. Conf. Com- put. Vis. Pattern Recognit., 2021, pp. 14 164–14 173

    2021

  29. [37]

    Incremental transformer structure enhanced image inpainting with masking positional encoding,

    Q. Dong, C. Cao, and Y . Fu, “Incremental transformer structure enhanced image inpainting with masking positional encoding,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2022, pp. 11 358–11 368

    2022

  30. [38]

    Keys to better image inpainting: Structure and texture go hand in hand,

    J. Jain, Y . Zhou, and N. Yu, “Keys to better image inpainting: Structure and texture go hand in hand,” in Winter Conf. Appl. Comput. Vis., 2023

    2023

  31. [39]

    Large scale image completion via co-modulated generative adversarial networks,

    S. Zhao, J. Cui, Y . Sheng, Y . Dong, X. Liang, E. I. Chang, and Y . Xu, “Large scale image completion via co-modulated generative adversarial networks,” Proc. Int. Conf. Learn. Representations, 2021

    2021

  32. [40]

    Image inpainting for irregular holes using partial convolutions,

    G. Liu, F. A. Reda, K. J. Shih, T.-C. Wang, A. Tao, and B. Catanzaro, “Image inpainting for irregular holes using partial convolutions,” inProc. Eur. Conf. Comput. Vis., 2018, pp. 85–100

    2018

  33. [41]

    Region normalization for image inpainting,

    T. Yu, Z. Guo, X. Jin, S. Wu, Z. Chen, W. Li, Z. Zhang, and S. Liu, “Region normalization for image inpainting,” in Proc. AAAI, 2020

    2020

  34. [42]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly et al., “An image is worth 16x16 words: Transformers for image recognition at scale,” Proc. Int. Conf. Learn. Representations, 2020

    2020

  35. [43]

    Diverse image inpainting with bidirectional and autoregressive transformers,

    Y . Yu, F. Zhan, R. Wu, J. Pan, K. Cui, S. Lu, F. Ma, X. Xie, and C. Miao, “Diverse image inpainting with bidirectional and autoregressive transformers,” in Proc. Int. Conf. Multimedia, 2021, pp. 69–78

    2021

  36. [44]

    Image inpainting guided by coherence priors of semantics and textures,

    L. Liao, J. Xiao, Z. Wang, C.-W. Lin, and S. Satoh, “Image inpainting guided by coherence priors of semantics and textures,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2021, pp. 6539–6548

    2021

  37. [45]

    Edgecon- nect: Generative image inpainting with adversarial edge learning,

    K. Nazeri, E. Ng, T. Joseph, F. Z. Qureshi, and M. Ebrahimi, “Edgecon- nect: Generative image inpainting with adversarial edge learning,” Proc. Int. Conf. Comput. Vis. Workshop, 2019

    2019

  38. [46]

    Contextual residual aggregation for ultra high-resolution image inpainting,

    Z. Yi, Q. Tang, S. Azizi, D. Jang, and Z. Xu, “Contextual residual aggregation for ultra high-resolution image inpainting,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2020, pp. 7508–7517

    2020

  39. [47]

    Unleashing transformers: Parallel token prediction with discrete ab- sorbing diffusion for fast high-resolution image generation from vector- quantized codes,

    S. Bond-Taylor, P. Hessey, H. Sasaki, T. P. Breckon, and C. G. Willcocks, “Unleashing transformers: Parallel token prediction with discrete ab- sorbing diffusion for fast high-resolution image generation from vector- quantized codes,” in Proc. Eur. Conf. Comput. Vis., 2022, pp...

    2022

  40. [48]

    Come-closer-diffuse-faster: Accelerat- ing conditional diffusion models for inverse problems through stochastic contraction,

    H. Chung, B. Sim, and J. C. Ye, “Come-closer-diffuse-faster: Accelerat- ing conditional diffusion models for inverse problems through stochastic contraction,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2022

    2022

  41. [49]

    Cold diffusion: In- verting arbitrary image transforms without noise,

    A. Bansal, E. Borgnia, H.-M. Chu, J. S. Li, H. Kazemi, F. Huang, M. Goldblum, J. Geiping, and T. Goldstein, “Cold diffusion: In- verting arbitrary image transforms without noise,” arXiv preprint arXiv:2208.09392, 2022

    2022 arXiv

  42. [50]

    Delving globally into texture and structure for image inpainting,

    H. Liu, Y . Wang, M. Wang, and Y . Rui, “Delving globally into texture and structure for image inpainting,” in Proc. ACM Int. Conf. Multimedia, 2022, pp. 1270–1278

    2022

  43. [51]

    Dual diffusion implicit bridges for image-to-image translation,

    X. Su, J. Song, C. Meng, and S. Ermon, “Dual diffusion implicit bridges for image-to-image translation,” Proc. Int. Conf. Learn. Represent., 2023

    2023

  44. [52]

    Slotdiffusion: Object-centric generative modeling with diffusion models,

    Z. Wu, J. Hu, W. Lu, I. Gilitschenski, and A. Garg, “Slotdiffusion: Object-centric generative modeling with diffusion models,” Proc. Ad- vances Neural Inf. Process. Syst., 2023

    2023

  45. [53]

    gddim: Generalized denoising diffusion implicit models,

    Q. Zhang, M. Tao, and Y . Chen, “gddim: Generalized denoising diffusion implicit models,” Proc. Int. Conf. Learn. Represent., 2023

    2023

  46. [54]

    Generative modeling by estimating gradients of the data distribution,

    Y . Song and S. Ermon, “Generative modeling by estimating gradients of the data distribution,” Proc. Advances Neural Inf. Process. Syst., 2019

    2019

  47. [55]

    Denoising diffusion restoration models,

    B. Kawar, M. Elad, S. Ermon, and J. Song, “Denoising diffusion restoration models,” Proc. Advances Neural Inf. Process. Syst., 2022

    2022

  48. [56]

    Palette: Image-to-image diffusion models,

    C. Saharia, W. Chan, H. Chang, C. Lee, J. Ho, T. Salimans, D. Fleet, and M. Norouzi, “Palette: Image-to-image diffusion models,” in Proc. ACM SIGGRAPH Conf., 2022, pp. 1–10. 14

    2022

  49. [57]

    Glide: Towards photorealistic image gen- eration and editing with text-guided diffusion models,

    A. Nichol, P. Dhariwal, A. Ramesh, P. Shyam, P. Mishkin, B. McGrew, I. Sutskever, and M. Chen, “Glide: Towards photorealistic image gen- eration and editing with text-guided diffusion models,” Proc. Int. Conf. Learn. Represent., 2021

    2021

  50. [58]

    Smartbrush: Text and shape guided object inpainting with diffusion model,

    S. Xie, Z. Zhang, Z. Lin, T. Hinz, and K. Zhang, “Smartbrush: Text and shape guided object inpainting with diffusion model,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2023, pp. 22 428–22 437

    2023

  51. [59]

    High- resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2022

    2022

  52. [60]

    Pseudoinverse-guided diffusion models for inverse problems,

    J. Song, A. Vahdat, M. Mardani, and J. Kautz, “Pseudoinverse-guided diffusion models for inverse problems,” in Proc. Int. Conf. Learn. Represent., 2023

    2023

  53. [61]

    I 2sb: Image-to-image schrodinger bridge,

    G.-H. Liu, A. Vahdat, D.-A. Huang, E. A. Theodorou, W. Nie, and A. Anandkumar, “I 2sb: Image-to-image schrodinger bridge,” Proc. Int. Conf. Mach. Learn., 2023

    2023

  54. [62]

    Direct diffusion bridge using data consistency for inverse problems,

    H. Chung, J. Kim, and J. C. Ye, “Direct diffusion bridge using data consistency for inverse problems,” Proc. Advances Neural Inf. Process. Syst., vol. 36, 2024

    2024

  55. [63]

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

    L. Zhang, A. Rao, and M. Agrawala, “Adding conditional control to text- to-image diffusion models,” in Proc. Int. Conf. Comput. Vis. , 2023, pp. 3836–3847

    2023

  56. [64]

    A task is worth one word: Learning with task prompts for high-quality versatile image inpainting,

    J. Zhuang, Y . Zeng, W. Liu, C. Yuan, and K. Chen, “A task is worth one word: Learning with task prompts for high-quality versatile image inpainting,” in Proc. Eur. Conf. Comput. Vis., 2025, pp. 195–211

    2025

  57. [65]

    Deep equilibrium approaches to diffusion models,

    A. Pokle, Z. Geng, and J. Z. Kolter, “Deep equilibrium approaches to diffusion models,” Proc. Advances Neural Inf. Process. Syst., vol. 35, pp. 37 975–37 990, 2022

    2022

  58. [66]

    Diffusion models beat gans on image synthesis,

    P. Dhariwal and A. Nichol, “Diffusion models beat gans on image synthesis,” Proc. Advances Neural Inf. Process. Syst., 2021

    2021

  59. [67]

    Statistical aspects of wasserstein dis- tances,

    V . M. Panaretos and Y . Zemel, “Statistical aspects of wasserstein dis- tances,” Annu. Rev. Stat. Appl., vol. 6, no. 1, pp. 405–431, 2019

    2019

  60. [68]

    Large-scale wasserstein gradient flows,

    P. Mokrov, A. Korotin, L. Li, A. Genevay, J. M. Solomon, and E. Bur- naev, “Large-scale wasserstein gradient flows,” Proc. Advances Neural Inf. Process. Syst., vol. 34, pp. 15 243–15 256, 2021

    2021

  61. [69]

    Variational wasserstein gradient flow,

    J. Fan, Q. Zhang, A. Taghvaei, and Y . Chen, “Variational wasserstein gradient flow,” Proc. Int. Conf. Mach. Learn., 2022

    2022

  62. [70]

    Deep learning face attributes in the wild,

    Z. Liu, P. Luo, X. Wang, and X. Tang, “Deep learning face attributes in the wild,” in Proc. Int. Conf. Comput. Vis., 2015, pp. 3730–3738

    2015

  63. [71]

    Places: A 10 million image database for scene recognition,

    B. Zhou, A. Lapedriza, A. Khosla, A. Oliva, and A. Torralba, “Places: A 10 million image database for scene recognition,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 40, no. 6, pp. 1452–1464, 2017

    2017

  64. [72]

    Imagenet large scale visual recognition challenge,

    O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla, M. Bernstein et al., “Imagenet large scale visual recognition challenge,” Int. J. Comput. Vis., vol. 115, no. 3, 2015

    2015

  65. [73]

    Ntire 2022 image inpainting challenge: Report,

    A. Romero, A. Castillo, J. Abril-Nova, R. Timofte, R. Das, S. Hira, Z. Pan, M. Zhang, B. Li, D. He et al. , “Ntire 2022 image inpainting challenge: Report,” in Proc. Conf. Comput. Vis. Pattern Recognit., 2022. Pourya Shamsolmoali (Senior Member, IEEE) received the PhD degree i...

    2022

  66. [2008]

    He received the DAGM Olympus award in 2005 and is fellow of the IAPR and ELLIS. His research interests include video object and instance segmentation, classification, segmen- tation, and registration of point clouds, as well as efficient machine learning techniques for in- cre...

    2005

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.