RenderMatte: Exact-Alpha Rendering and Group-Relative Alignment for Image Matting

Jianing Zhao; Qun Jin; Ruichen Cong; Yafei Hu; Yiren Song; Zecheng Ren

REVIEW 5 major objections 5 minor 51 references

RenderMatte: Exact-Alpha Rendering and Group-Relative Alignment for Image Matting

T0 review · 5 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read RenderMatte claims that trimap-guided image matting can be recast as a structure-preserving image-editing task, and that exact-alpha synthetic data, an alpha-edge loss, and group-relative reward alignment together push alpha prediction…

desk verdict Solid method and dataset, but the reported SOTA is undermined by test-set checkpoint selection and unmatched baselines; worth a revision, not a desk reject. read the letter →

arxiv 2608.08487 v1 pith:MWMJJSKL submitted 2026-08-09 cs.CV

Zecheng Ren , Yafei Hu , Jianing Zhao , Ruichen Cong , Qun Jin , Yiren Song This is my paper · ORCID

classification cs.CV

keywords imagemattingtrimap-guidedalphamatteflowmatchingeditingpriorsgroup-relativealignmentsyntheticdatasetdiffusiontransformer

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

RenderMatte tries to establish that image matting—estimating a continuous per-pixel opacity value so a foreground can be cleanly separated from its background—can be solved by treating it as a conditional image-editing problem rather than a dedicated segmentation task. The paper fine-tunes the FLUX.1 Kontext image editor to output alpha mattes instead of RGB images, adds a pixel-space alpha-edge loss that sharpens sparse boundaries such as hair and fur, and then post-trains the model by sampling several mattes per input and ranking them with matting-specific rewards. To support this, it builds an 83,533-image synthetic dataset whose alpha annotations come from 3D-rendered, GPT-generated, and Internet-collected RGBA foregrounds, including exact strand-level boundaries. If the claim is right, large image-editing priors can be transferred to precise fractional-opacity prediction, and the reported results—best average rank of 1.6 across AIM-500, P3M-500-NP, AM-2K, and RenderMatte-2K—suggest a scalable route to open-world matting.

What carries the argument

The load-bearing mechanism is a two-stage training procedure built on the FLUX.1 Kontext flow-matching editor. The diffusion-transformer (DiT) backbone is fully fine-tuned to map a concatenation of input image, trimap, and text tokens to a matte-shaped target latent, using the rectified-flow objective $\mathcal{L}_{FM} = \mathbb{E}\|v_\theta(\mathrm{concat}(z_t, c_{x,g}, c_p), t) - (z_1-z_0)\|_2^2$, plus an $\alpha$-edge loss $\mathcal{L}_{AE}$ that applies an $\ell^1$ error separately on uncertain and certain trimap regions and a multi-scale Laplacian-pyramid boundary term on the uncertain band. After supervised fine-tuning, a group-relative alignment stage samples $G=8$ candidate mattes per condition, scores them with a terminal reward $R_i = -\sum_{m} w_m \min(e_m(\hat{y}_i,y,g)/s_m, \tau)$, standardizes the rewards into advantages, and optimizes a clipped GRPO-style objective through a rank-64 LoRA adapter while the base model stays frozen.

What would settle it

Train or evaluate the full pipeline on a held-out set of real photographs with manual, strand-accurate alpha ground truth and compare boundary metrics (Grad, Conn) against a model trained only on real composites; if the synthetic-trained model's boundary errors approach or exceed those of a non-generative baseline trained on real data, the transfer assumption fails. A cheaper test: ablate the dataset by source (3D-rendered vs. GPT-generated vs. Internet-collected) and see whether boundary metrics collapse when 3D-rendered examples are removed.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the hardest part of matting—high-frequency, semi-transparent boundary structure—responds to a combination of exact alpha supervision and relative reward alignment, not to any single loss. Repurposing FLUX.1 Kontext's editing prior through full-parameter fine-tuning gives the model a strong semantic and structural initialization. The alpha-edge loss, which separates the uncertain trimap band from known regions and adds a Laplacian-pyramid boundary term, recovers the sparse strands that latent flow matching alone smooths; the ablation attributes the clearest gains in Grad and SAD to this term. Group-relative alpha alignment then treats each of eight sampled mattes as a competing candidate and standardizes an alpha-space reward—covering alpha accuracy, boundary fidelity, trimap compliance, and compositional consistency—into advantages, giving a refinement signal that continues to improve the reward curve after supervised adaptation has saturated. The paper reports that the combined pipeline reaches an average rank of 1.6 and the lowest errors on most metrics across the four benchmarks.

Load-bearing premise

The load-bearing assumption is that the 83,533 fully synthetic composites reproduce the alpha distributions, boundary statistics, and appearance of real photographs closely enough that gains learned on them transfer to real-world matting benchmarks; the paper does not analyze the domain gap or ablate training-data composition.

Editorial extensions

If this is right

If the central claim is right, generative image editors can be repurposed as dense structure-preserving predictors: the same backbone that edits photos can be fine-tuned to output alpha mattes without custom matting architectures.
Exact strand-level alpha supervision from 3D rendering is sufficient at scale: 83,533 composites built from rendered, generated, and collected RGBA foregrounds yield zero-shot performance on real benchmarks.
The two-stage design separates representation learning from quality alignment: the alpha-edge loss supplies boundary sharpness, while group-relative reward alignment supplies a refinement stage that continues to improve after supervised losses saturate.
The pipeline trades inference speed for accuracy: about 2.7 seconds per 1024x1024 image on an A100, versus under 0.5 seconds for DiffMatte, which constrains interactive use until efficiency work is done.
The reward curve reaching its best value at step 256 indicates that reward-driven exploration adds information beyond supervised adaptation, justifying group-relative alignment as a post-training stage.

Reading between the lines

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

A natural extension is to apply the same group-relative reward alignment to other spatially aligned dense-prediction targets, such as depth, surface normals, or segmentation, where multiple sampled outputs can be ranked by task-specific rewards; the paper's alpha-space reward is already architecture-agnostic.
The fixed text instruction—'Transform to matting map while maintaining original composition'—means semantic conditioning is carried almost entirely by the visual tokens; testing variable prompts would clarify how much of FLUX.1 Kontext's editing prior is genuinely semantic rather than structural.
If inference latency is the bottleneck, a one-step or distilled variant of the flow-matching trajectory could preserve most boundary gains while closing the roughly 2.2-second gap versus DiffMatte; the paper leaves this open.
The synthetic-data assumption could be probed by measuring per-source generalization: 3D-rendered assets provide exact boundaries but their appearance may be less diverse than Internet-collected assets, so ablating source proportions would show which part of the data actually drives the boundary improvements.

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.

Desk Editor's Note

read the letter

Two things worth knowing. The paper builds a genuinely new training recipe for matting — adapting FLUX.1 Kontext with an alpha-edge pixel-space loss and GRPO-style reward alignment over sampled mattes — and it contributes a new synthetic dataset with exact alpha ground truth. That combination is new, and the dataset is a real asset. The second thing is less good: the headline 'state-of-the-art across all benchmarks' is not supported by the experiments as reported, because the checkpoints were selected using the same test benchmarks that Table 1 is built on. The appendix confirms the SFT run computed full metrics on AIM-500, P3M-500-NP, and AM-2K at every saved checkpoint and chose step 10,000 based on that comparison; the alignment stage then uses the AIM-500 reward curve to pick step 256 as best. The resulting Table 1 numbers are a best-of-several selection on the test set, with no error bars or repeated runs. On top of that, the baselines look copied from earlier papers without a statement about trimap dilation, resolution, or sampling protocol. That gives the evaluation a selection-bias and protocol-mismatch smell, and the 1.6 average rank should not be taken at face value until a selection-free rerun is done and baselines are re-run under the same conditions.

The method itself deserves credit. The alpha-edge loss is a sensible way to strengthen boundary supervision in pixel space over latent flow matching, and the reward function is plausibly designed for matting, though the exact weights, scales, and threshold aren't specified. The ablation in Table 2 shows a clean monotonic improvement from base model to alpha-edge to reward alignment, which is internally consistent. The limitations section is honest about inference cost and OOD failures. The citation pattern is fine — DiffMatte and Edit2Perceive are credited, and the GRPO lineage is explicit. The synthetic-to-real transfer assumption is real but secondary — their own RenderMatte-2K benchmark addresses zero-shot generalization, and a controlled comparison on public benchmarks would answer the transfer question directly.

Who is this for? Matting researchers and people adapting generative editors to dense prediction. They get a useful data recipe and a training paradigm, plus a reminder about evaluation hygiene. I would not cite the SOTA numbers yet, but I would cite the dataset and method if the code and data are released as promised.

Recommendation: send it to peer review with a clear request for a selection-free evaluation and protocol-matched baselines. The core ideas are viable; the evidence needs to be made honest.

Referee Report

5 major / 5 minor

Summary. The paper introduces RenderMatte, a trimap-guided image matting framework that adapts the FLUX.1 Kontext image-editing model through full-parameter fine-tuning. The training pipeline combines latent flow matching with an alpha-edge pixel-space loss, followed by a group-relative alpha alignment stage that uses a GRPO-style clipped policy objective with matting-specific rewards. To provide exact alpha supervision at scale, the authors construct the RenderMatte dataset, containing 83,533 composites from 3D-rendered, GPT-generated, and Internet-collected RGBA foregrounds, and evaluate on AIM-500, P3M-500-NP, AM-2K, and a new RenderMatte-2K benchmark. The paper reports the best average rank of 1.6 and claims state-of-the-art performance across all benchmarks.

Significance. If the reported results withstand scrutiny, this is a substantial contribution: it demonstrates that a large image-editing prior can be adapted to alpha prediction, that pixel-space boundary supervision complements latent flow matching, and that reward-based alignment with dense matting metrics yields further gains. The release of the synthetic dataset with exact alpha annotations, code, and model checkpoints would be valuable for the community. The main significance is the combination of these three ideas and the scale of synthetic data. However, the current evidence for the SOTA claim is weakened by checkpoint selection on the evaluation benchmarks, an underspecified reward function, and limited analysis of the synthetic-to-real transfer, so the contribution cannot yet be endorsed at face value.

major comments (5)

[Appendix, SFT Checkpoint Selection and Figure 5] The paper performs model selection on the evaluation benchmarks themselves. The appendix states that the full 500-sample AIM-500 evaluation is run at each saved SFT checkpoint and that step 10,000 was selected after comparing complete metrics on AIM-500, P3M-500-NP, and AM-2K (Table 4). These are the same benchmarks reported as test sets in Table 1. In addition, the alignment stage is evaluated on AIM-500 and Figure 5 singles out step 256 as the best reward, but the text does not disclose which alignment checkpoint produced Table 1. This protocol makes the 'zero-shot' characterization misleading and is likely to inflate the reported numbers. Please provide a selection-free evaluation (for example, a fixed checkpoint chosen on a separated validation split) and report the full checkpoint curves for all benchmarks.
[Table 1 and Implementation Details] The comparison protocol for baseline methods is not specified. The paper does not state whether baseline numbers were re-run under identical trimap dilation, resolution, and metric evaluation settings, or whether they were copied from prior publications. The dagger markers for SDMatte and Edit2Perceive indicate a different interaction format, but they do not make the evaluation protocol transparent. Please specify the trimap generation parameters for all methods, the image resolution used, the exact metric implementation, and whether any baselines are taken from the literature; otherwise the claimed improvements over methods such as ViTMatte and DiffMatte are not fully interpretable.
[Equation (9), Matting reward] The reward function that is central to the group-relative alignment contribution is underspecified. The sets of error criteria, the weights w_m, the normalization scales s_m, and the clipping threshold tau are not defined or given numerical values, and the four error families (alpha accuracy, boundary fidelity, structural consistency, trimap compliance) are only named informally. Without this information the alignment stage cannot be reproduced, and the claim that this reward drives the observed improvement is not testable. Please provide the complete reward specification and include an ablation of the reward hyperparameters (weights, scales, threshold).
[Large-Scale Synthetic Matting Data / Datasets and Benchmarks] The model is trained exclusively on the synthetic RenderMatte dataset, and the paper offers no direct analysis of the synthetic-to-real domain gap. The claim that exact synthetic alpha supervision transfers to real-world matting benchmarks is the central empirical bridge of the paper, yet there is no ablation varying the training-data composition (for example, removing the GPT-generated, Internet, or 3D-rendered sources) and no quantitative comparison of synthetic versus real alpha statistics. Such an analysis would substantially strengthen the generalization claim and should be added.
[Table 1 and Evaluation Metrics] All metrics in Table 1 are reported from a single run with no variance estimates. Given that the reported differences on P3M-500-NP and AM-2K are very small (for example, MSE values of 0.001 for RenderMatte, ViTMatte, and DiffMatte, and MAD values of 0.004 for several methods), the claimed improvement could be within run-to-run variability. Please report repeated runs with standard deviations or, at minimum, bootstrap confidence intervals over the evaluation set.

minor comments (5)

[Abstract and Conclusion] The phrase 'state-of-the-art performance across all benchmarks' is stronger than what Table 1 supports, since RenderMatte does not achieve the lowest value on every metric and several values are tied with baselines; please moderate the wording.
[Figure 2 and Method] The figure and text refer to 'SDE Sampling', but the stochastic differential equation, the noise schedule, and the exploration window are not defined until the appendix; please introduce the SDE formalism in the Method section or simplify the terminology.
[Equation (3)] The notation 1/|Omega_k| is ambiguous because the 'certain' region contains both definite foreground and definite background, and the two are likely to be weighted differently; please clarify the partition and weighting.
[Alpha-Edge loss] The term L_bd is described as a multi-scale Laplacian pyramid loss, but no explicit definition is given beyond the reference to Burt and Adelson; please provide the exact formula or a clear algorithmic description.
[Group-Relative Alpha Alignment] The text uses 'unknown region' and 'uncertain region' interchangeably; please unify the terminology, as this appears in Eq. (3) and the surrounding discussion.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation; appendix test-set checkpoint selection is a validity risk, not a circularity.

full rationale

The training and post-training objectives are defined on ground-truth alpha from synthetic composites: the flow-matching loss (Eq. 2), the alpha-edge loss (Eq. 3), and the matting reward (Eq. 9) all take y and g as inputs, never the public benchmark numbers. Evaluation is on external benchmarks (AIM-500, P3M-500-NP, AM-2K) plus a held-out synthetic benchmark, so the reported numbers are measurements of a trained model rather than re-statements of the loss definitions. The reward's use of the same error family as the evaluation metrics is objective alignment, not a circular derivation. The self-citations to DiffMatte and Edit2Perceive are motivational and are not load-bearing: they do not supply a uniqueness theorem, forbid alternatives, or enter the equations. The appendix does disclose a protocol risk: 'the full 500-sample AIM-500 evaluation is run at each saved checkpoint' and 'Table 4 compares the checkpoints with complete benchmark metrics,' with the AIM-500 reward curve reaching 'its best value at step 256' (Fig. 5/Table 5). If the final Table 1 checkpoint was chosen from these evaluations, the SOTA claim could be inflated by test-set selection, and baseline protocol matching is not demonstrated. That is a correctness/transparency concern, not a circularity under the strict definition: no fitted parameter is renamed as a prediction through an equation, and the final checkpoint is not explicitly identified as the test-selected best. Hence score 2 rather than 0, reflecting the non-load-bearing self-citations and the need to disclose the selection protocol.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The ledger is small because this is an empirical ML paper. The main hand-tuned elements are the reward and loss weights, which are not disclosed, and the main assumptions concern synthetic-to-real transferability. No new physical or theoretical entities are introduced.

free parameters (3)

Reward criterion weights w_m, scales s_m, and clipping threshold tau in Eq. (9) = not reported
These control the reward that drives group-relative alignment; without them the post-training objective is undefined, and the reported results depend on their values.
Alpha-edge loss weight lambda_bd and schedule parameters Niter, e0 in Eqs. (3)-(5) = not reported
lambda_bd weights boundary supervision; e0 controls when the pixel-space objective switches on. Values are hand-chosen and not specified in the paper.
Group size G=8, clipping epsilon=1e-4, noise level 0.2, guidance scale 1.0, LoRA rank 64, learning rates = given in implementation details
Key hyperparameters of the alignment stage; the paper reports them but does not justify them, and the SOTA outcome is conditional on these choices.

assumptions (5)

domain assumption Alpha compositing (Porter-Duff) plus a per-pixel opacity model is a valid generative model for matting training data.
In the dataset section, composite images are generated with I = alpha F + (1-alpha) B, which assumes linear compositing.
domain assumption FLUX.1 Kontext's learned editing prior transfers to alpha prediction when fine-tuned on RGBA composites.
The whole method relies on this transfer; no analysis directly measures the contribution of the editing prior vs. the synthetic data.
domain assumption The RenderMatte synthetic distribution is representative of real-world matting test sets.
Entry point of the weakest assumption; the paper evaluates on real benchmarks without a domain-gap analysis.
ad hoc to paper The reward function defined in Eq. (9), with unspecified weights and scales, is a good proxy for human-judged matte quality.
The reward terms are chosen by the authors and are not derived from user studies.
standard math Standard mathematical tools: rectified flow, morphological trimap dilation and erosion, Laplacian pyramid.
Background machinery used without proof in the paper.

how reviews work

0 comments

Cite this review

Pith. "Pith review of RenderMatte: Exact-Alpha Rendering and Group-Relative Alignment for Image Matting." pith.science (2026). https://pith.science/paper/MWMJJSKL

@misc{pith2026260808487,
  author       = {Pith},
  title        = {Pith review of: RenderMatte: Exact-Alpha Rendering and Group-Relative Alignment for Image Matting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MWMJJSKL}},
  note         = {Machine review of arXiv:2608.08487}
}

read the original abstract

Image matting is an essential enabling technology for modern visual content production, where foreground extraction determines the realism and editability of downstream creation workflows. However, precise alpha estimation in open-world scenes remains challenging because real foregrounds exhibit highly diverse appearances and opacity patterns. This makes existing methods struggle with semantic ambiguity and fine-grained opacity variation, especially in sparse boundary regions that are fragile and difficult to supervise. To address this gap, we present RenderMatte, a trimap-guided matting framework that adapts FLUX.1 Kontext through full-parameter fine-tuning, leveraging image editing priors for structure-preserving alpha prediction. During supervised adaptation, an alpha-edge objective preserves the latent flow-matching signal while strengthening pixel-space boundary supervision. We further introduce group-relative alpha alignment for post-training. It compares multiple mattes sampled under the same trimap condition using matting-specific rewards for alpha accuracy, boundary fidelity, trimap compliance, and compositional consistency. To overcome the lack of precise edge annotations, we construct the RenderMatte dataset, a large-scale synthetic dataset combining 3D-rendered RGBA foregrounds with diverse multi-source assets. It features exact strand-level alpha annotations and diverse background composites. Experiments show state-of-the-art performance across all benchmarks, demonstrating a scalable path toward high-fidelity matting in open-world scenes.

Figures

Figures reproduced from arXiv: 2608.08487 by the authors.

**Figure 1.** We present RenderMatte, a trimap-guided image matting framework built upon image editing priors. Our model achieves state-of-the-art performance across the zero-shot matting task, consistently outperforming previous methods. Radar chart summarizing average rank performance across four benchmarks; points closer to the outer edge indicate superior performance. Zoom in for better view. Abstract Image matting is an ess… view at source ↗

**Figure 2.** Overview of the RenderMatte framework. We adapt the FLUX.1 Kontext editor for image matting (Labs et al. 2025). Given a text prompt p, target image y, and input image x, the model predicts an alpha matte yˆ. In the forward process, the target alpha token is encoded as z1, noised into zt, and concatenated with the text and image tokens. The trainable DiT backbone predicts the velocity vθ from noise z0 to the target l… view at source ↗

**Figure 3.** Overview of RenderMatte dataset construction. (a) RenderMatte contains 83,533 composite training images from three foreground sources: 3D-rendered assets, GPT-generated assets, and Internet-collected assets. (b) We synthesize training samples by pairing each foreground with a background image. This pipeline preserves exact alpha annotations while increasing foreground–background diversity. (c) For 3D assets, we rend… view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Qualitative comparison with state-of-the-art image matting methods. From left to right, we show the input image, [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: Evaluation reward curve of group-relative alpha [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 6.** Figure 6: Visual ablation of the proposed training stages. From left to right, we show the input image, trimap, predictions from [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

**Figure 7.** Figure 7: Failure cases on out-of-domain data. Top to bottom: input, RenderMatte, and ground truth. The model may produce [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 27 canonical work pages

[1]

Advances in neural information processing systems , volume=

Denoising diffusion probabilistic models , author=. Advances in neural information processing systems , volume=
[2]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

High-resolution image synthesis with latent diffusion models , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[3]

The eleventh international conference on learning representations , year=

Flow matching for generative modeling , author=. The eleventh international conference on learning representations , year=
[4]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Instructpix2pix: Learning to follow image editing instructions , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[5]

Proceedings of the IEEE/CVF international conference on computer vision , pages=

Adding conditional control to text-to-image diffusion models , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=
[6]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

FE2E: From Editor to Dense Geometry Estimator , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

work page
[7]

arXiv preprint arXiv:2511.18673 , year=

Edit2Perceive: Image Editing Diffusion Models Are Strong Dense Perceivers , author=. arXiv preprint arXiv:2511.18673 , year=

arXiv
[8]

and Lischinski, D

Levin, A. and Lischinski, D. and Weiss, Y. , booktitle=. A Closed Form Solution to Natural Image Matting , year=

work page

Show all 51 references

[9]

IEEE transactions on pattern analysis and machine intelligence , volume=

KNN matting , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2013 , publisher=

2013
[10]

Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

Deep image matting , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
[11]

Proceedings of the AAAI conference on artificial intelligence , volume=

Natural image matting via guided contextual attention , author=. Proceedings of the AAAI conference on artificial intelligence , volume=
[12]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Matteformer: Transformer-based image matting via prior-tokens , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[13]

Information Fusion , volume=

Vitmatte: Boosting image matting with pre-trained plain vision transformers , author=. Information Fusion , volume=. 2024 , publisher=

2024
[14]

Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =

Huang, Longfei and Liang, Yu and Zhang, Hao and Chen, Jinwei and Dong, Wei and Chen, Lunde and Liu, Wanyu and Li, Bo and Jiang, Peng-Tao , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2025 , pages =

2025
[15]

arXiv preprint arXiv:2305.13301 , year=

Training diffusion models with reinforcement learning , author=. arXiv preprint arXiv:2305.13301 , year=

arXiv
[16]

Advances in Neural Information Processing Systems , volume=

Dpok: Reinforcement learning for fine-tuning text-to-image diffusion models , author=. Advances in Neural Information Processing Systems , volume=
[17]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Diffusion model alignment using direct preference optimization , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[18]

arXiv preprint arXiv:2505.07818 , year=

DanceGRPO: Unleashing GRPO on Visual Generation , author=. arXiv preprint arXiv:2505.07818 , year=

arXiv
[19]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Mask-guided matting in the wild , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[20]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Ultrahigh resolution image/video matting with spatio-temporal sparsity , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[21]

Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

Trans-adapter: A plug-and-play framework for transparent image inpainting , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
[22]

arXiv preprint arXiv:2512.15603 , year=

Qwen-image-layered: Towards inherent editability via layer decomposition , author=. arXiv preprint arXiv:2512.15603 , year=
[23]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Unifying automatic and interactive matting with pretrained vits , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[24]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Magick: A large-scale captioned dataset from matting generated images using chroma keying , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[25]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

Chen, Xinyi and Dong, Hang and Jiang, Baowei and Xu, Shenkun and Guan, Youqi and Shi, Kanle and Gai, Kun and Song, Haichuan , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2026 , pages =

2026
[26]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Repurposing diffusion-based image generators for monocular depth estimation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[27]

European Conference on Computer Vision , pages=

Diffusion for natural image matting , author=. European Conference on Computer Vision , pages=. 2024 , organization=

2024
[28]

ACM SIGGRAPH 2024 Conference Papers , pages=

Matting by generation , author=. ACM SIGGRAPH 2024 Conference Papers , pages=

2024
[29]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

In-context matting , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[30]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Aesthetic post-training diffusion models from generic preferences with step-by-step preference optimization , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[31]

arXiv preprint arXiv:2107.07235 , year=

Deep automatic natural image matting , author=. arXiv preprint arXiv:2107.07235 , year=

arXiv
[32]

Proceedings of the 29th ACM international conference on multimedia , pages=

Privacy-preserving portrait matting , author=. Proceedings of the 29th ACM international conference on multimedia , pages=
[33]

International Journal of Computer Vision , volume=

Bridging composite and real: towards end-to-end deep image matting , author=. International Journal of Computer Vision , volume=. 2022 , publisher=

2022
[34]

Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

Indices matter: Learning to index for deep image matting , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
[35]

Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

Context-Aware Image Matting for Simultaneous Foreground and Alpha Estimation , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
[36]

2024 , eprint=

Aligning Text-to-Image Diffusion Models with Reward Backpropagation , author=. 2024 , eprint=

2024
[37]

1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space , author=

FLUX. 1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space , author=. arXiv preprint arXiv:2506.15742 , year=

arXiv
[38]

International conference on learning representations (ICLR) , year=

Flow straight and fast: Learning to generate and transfer data with rectified flow , author=. International conference on learning representations (ICLR) , year=
[39]

Proceedings of the IEEE/CVF international conference on computer vision , pages=

Scalable diffusion models with transformers , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=
[40]

Readings in computer vision , pages=

The Laplacian pyramid as a compact image code , author=. Readings in computer vision , pages=. 1987 , publisher=

1987
[41]

Proceedings of the AAAI conference on artificial intelligence , volume=

Modnet: Real-time trimap-free portrait matting via objective decomposition , author=. Proceedings of the AAAI conference on artificial intelligence , volume=
[42]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Kubric: A scalable dataset generator , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[43]

Proceedings of the IEEE/CVF international conference on computer vision , pages=

Hypersim: A photorealistic synthetic dataset for holistic indoor scene understanding , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=
[44]

arXiv preprint arXiv:2212.08051 , year=

Objaverse: A universe of annotated 3d objects , author=. arXiv preprint arXiv:2212.08051 , year=

arXiv
[45]

arXiv preprint arXiv:2204.06125 , volume=

Hierarchical text-conditional image generation with clip latents , author=. arXiv preprint arXiv:2204.06125 , volume=

arXiv
[46]

Advances in Neural Information Processing Systems , volume=

Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding , author=. Advances in Neural Information Processing Systems , volume=
[47]

International Conference on Learning Representations , volume=

Sdxl: Improving latent diffusion models for high-resolution image synthesis , author=. International Conference on Learning Representations , volume=
[48]

Proceedings of the 11th annual conference on Computer graphics and interactive techniques , pages=

Compositing digital images , author=. Proceedings of the 11th annual conference on Computer graphics and interactive techniques , pages=
[49]

Proceedings of the 26th ACM international conference on Multimedia , pages=

Semantic human matting , author=. Proceedings of the 26th ACM international conference on Multimedia , pages=
[50]

Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=

Robust high-resolution video matting with temporal guidance , author=. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=
[51]

arXiv preprint arXiv:2003.07711 , year=

F , B , Alpha Matting , author=. arXiv preprint arXiv:2003.07711 , year=

2003 arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Advances in neural information processing systems , volume=

Denoising diffusion probabilistic models , author=. Advances in neural information processing systems , volume=

[2] [2]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

High-resolution image synthesis with latent diffusion models , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

[3] [3]

The eleventh international conference on learning representations , year=

Flow matching for generative modeling , author=. The eleventh international conference on learning representations , year=

[4] [4]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Instructpix2pix: Learning to follow image editing instructions , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

[5] [5]

Proceedings of the IEEE/CVF international conference on computer vision , pages=

Adding conditional control to text-to-image diffusion models , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

[6] [6]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

FE2E: From Editor to Dense Geometry Estimator , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

work page

[7] [7]

arXiv preprint arXiv:2511.18673 , year=

Edit2Perceive: Image Editing Diffusion Models Are Strong Dense Perceivers , author=. arXiv preprint arXiv:2511.18673 , year=

arXiv

[8] [8]

and Lischinski, D

Levin, A. and Lischinski, D. and Weiss, Y. , booktitle=. A Closed Form Solution to Natural Image Matting , year=

work page