Pith. sign in

REVIEW 5 minor 50 references

From RGB Generation to Dense Field Readout: Pixel-Space Dense Prediction with Text-to-Image Models

T0 review · 0 major / 5 minor · reviewed 2026-07-11 · grok-4.5

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

Pith's one-line read Dense prediction can reuse a text-to-image model by reading task fields from its token lattice, not by generating image-like targets.

desk verdict Clean, well-supported interface paper: drop the target-side VAE and read dense fields linearly from a DiT lattice; SOTA and speed gains look real under matched ablations. read the letter →

arxiv 2607.06553 v2 pith:RXMKZEHT submitted 2026-07-07 cs.CV

Zanyi Wang , Xin Lin , Haodong Li , Dengyang Jiang , Yijiang Li , Pengtao Xie This is my paper · ORCID
classification cs.CV
keywords densepredictiontext-to-imagemodelsDiTReChannelpixel-spacereadoutgenerativepriorsLoRAmatting
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

Large text-to-image models already turn an RGB image into a patch-aligned grid of tokens. The authors argue that dense prediction does not need the rest of the generative pipeline: you do not have to encode depth, mattes, masks, or heatmaps into an RGB-trained latent space and decode them back as if they were new images. Their method, ReChannel, keeps the input encoder, freezes the transformer, adapts it with lightweight task LoRA, and maps each token to its local pixel patch with a shared linear head of roughly 33K parameters and no spatial mixing. Across six dense tasks and more than a dozen benchmarks this minimal readout sets new state of the art on trimap-free matting, KITTI depth, and referring segmentation, stays competitive elsewhere, and in a matched setting is both more accurate and 2.48 times faster than an edit-plus-latent-decode counterpart. A reader who cares about perception backbones gets a concrete claim: generative pretraining is useful for the field it organizes, not for the rendering interface it was trained to use.

What carries the argument

ReChannel: after task LoRA adapts the frozen DiT, a shared token-local linear head maps each spatial token to its fixed p by p by K_t pixel patch, reinterpreting the token's channels from RGB appearance to task-native fields with no target-side VAE and no inter-token mixing in the head.

What would settle it

On a matched backbone and resolution, if a larger spatial decoder or a target-side VAE or edit path clearly beat the token-local linear head on high-frequency matting or multi-channel pose heatmaps, the claim that the adapted token field already carries the output would fail.

Watch

Extended reading notes

Core claim

Dense perception can benefit from generative pretraining without inheriting its output interface. A pretrained DiT already organizes RGB inputs through a patch-to-token-to-patch lattice on the image plane, so each token indexes a fixed output patch whose channels can carry task-native quantities instead of RGB appearance. ReChannel therefore keeps the VAE encoder for the DiT's input distribution, drops the target-side decoder, adapts the frozen DiT with task LoRA, and reads each token to its p by p by K_t pixel-space patch through a shared token-local linear head.

Load-bearing premise

Once the transformer is lightly adapted for a task, its token grid already holds the spatial structure needed, so a purely local linear map with no mixing between tokens is enough for every dense target tried.

Editorial extensions

If this is right

  • Dense targets need not be forced through an RGB-trained VAE encode-decode path to reuse a text-to-image prior.
  • One frozen T2I backbone plus task LoRA and a shared linear readout can cover continuous geometry, alpha mattes, language-conditioned masks, and multi-channel heatmaps.
  • Removing the target-side generative path can improve both accuracy and latency (up to 2.48 times faster) versus edit-plus-latent-decode.
  • Extra head capacity or full fine-tuning need not beat a token-local linear readout once the token field is task-adapted.
  • New state-of-the-art results on trimap-free matting, KITTI depth, and referring segmentation are reachable with this interface.

Reading between the lines

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

  • If adapted token lattices already carry spatial structure, many decoder-heavy dense pipelines may be oversized once a strong generative prior is in place.
  • The same rechanneling pattern should transfer to other patch-token generative backbones, not only the family evaluated here.
  • Non-image-like but still pixel-aligned fields such as optical flow or multi-frame disparity stacks are natural next targets for the same linear readout form.
  • Discriminative foundation encoders might also admit similarly minimal readouts if light adaptation collapses their token fields into compact task subspaces.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 5 minor

Summary. The paper argues that dense prediction should not inherit the target-side VAE/edit interface of text-to-image models. A pretrained DiT already organizes RGB inputs on a patch o token o patch lattice, so each token can be read out as a fixed p imes p imes K_t task-native pixel patch rather than RGB appearance. ReChannel keeps the VAE encoder for the DiT input distribution, freezes the backbone, adapts with task LoRA, and uses a shared token-local linear head (~33K parameters, no spatial mixing). On FLUX-Klein (4B/9B) this recipe is evaluated on six dense tasks and a dozen-plus benchmarks, claiming new SOTA on trimap-free matting, KITTI depth, and referring segmentation, competitive results on normals/saliency/pose, and 2.48 imes lower latency than a matched edit-plus-latent-decode baseline under the same 4B backbone.

Significance. If the results hold, the work cleanly separates generative pretraining priors from the generative output interface and shows that a minimal token-local linear readout is sufficient for heterogeneous dense targets. The matched 4B ablations (Table 5) isolate LoRA necessity, pretraining necessity, head capacity, and generative-interface alternatives, while main tables report multi-benchmark gains and direct latency. This is a practical, falsifiable contribution for the growing line of generative dense predictors and is supported by public-benchmark protocols and a public code link. Strengths include the multi-task scope, the controlled interface ablations, and the explicit efficiency comparison under a fixed backbone.

minor comments (5)
  1. Fig. 2 participation-ratio values are labeled “illustrative”; a short appendix note on how PR is computed (token covariance, layer, after/before LoRA) would make the diagnostic fully reproducible.
  2. Sec. 3.3 / Eq. (2): state the concrete p and token dimension used for FLUX-Klein so the ~33K parameter count can be verified without external lookup.
  3. Table 1: ReChannel-9B trails Edit2Perc on NYU absRel (0.051 vs 0.044); a one-sentence discussion of this split would balance the SOTA claims on KITTI/ScanNet.
  4. Sec. 4.1: briefly list LoRA rank, learning rate, and epochs (or point to a config) so the matched-4B ablations can be re-run exactly.
  5. Conclusion already notes scope limits (backbone family, non-pixel-aligned targets); a short remark on resolution scaling of the pure token-local head would further clarify generality.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: empirical method paper evaluated on external public benchmarks with matched ablations.

full rationale

This is a standard empirical computer-vision paper proposing ReChannel (frozen DiT + task LoRA + shared token-local linear head mapping each token to a p imes p imes K_t pixel patch). All load-bearing claims are performance numbers on public datasets (NYU/KITTI/ScanNet depth, DSINE normals, P3M/AIM matting, RefCOCO-family referring segmentation, COCO pose, DUTS/ECSSD saliency) under each benchmark’s canonical protocol, plus controlled ablations (Table 5) that isolate head capacity, full fine-tuning, latent/VAE/edit interfaces, and random initialization under a fixed 4B backbone. No equation or derivation asserts a first-principles prediction that reduces by construction to a fitted parameter; no uniqueness theorem is imported from overlapping authors; self-citations are ordinary prior-work references, not load-bearing. The central interface claim is therefore externally falsifiable and independent of any circular step.

Assumptions & free parameters 3 free parameters · 3 assumptions · 2 invented entities

Empirical CV method paper. Load-bearing content is architectural choices and the empirical claim that a minimal linear readout suffices; almost everything else is standard DiT/VAE/LoRA machinery or conventional dense-prediction losses and protocols.

free parameters (3)
  • LoRA rank / scale per task
    Controls capacity of the task adapter; exact rank not stated in the main text yet the claim that lightweight adaptation is sufficient depends on it.
  • token-local linear head weights (~33K parameters)
    Fitted per task; the central sufficiency claim is that this tiny head is enough once the token field is adapted.
  • patch size p and output channels K_t
    Tied to the backbone’s patch lattice and task representation; chosen by design rather than free fit, but still free architectural choices that define the readout.
assumptions (3)
  • domain assumption A pretrained DiT organizes an RGB input into a fixed patch→token→patch spatial lattice on the image plane.
    Taken from the standard ViT/DiT architecture (Sec. 3); used as the substrate that makes token-local readout possible.
  • domain assumption RGB generation pretraining supplies useful semantic, structural and geometric priors for dense prediction.
    Inherited from prior generative dense-prediction literature (Marigold, Lotus, etc.) and confirmed by the random-init ablation in Table 5.
  • ad hoc to paper After LoRA adaptation the token field already contains the spatial structure of the dense target, so no spatial mixing in the head is required.
    Central empirical hypothesis of Sec. 3.3 and the diagnostic block of Table 5; not a mathematical necessity.
invented entities (2)
  • ReChannel (token-local linear field readout)
    purpose: Name and concrete recipe for reinterpreting DiT token channels as task-native pixel patches without a target-side VAE decoder.
    The paper’s primary methodological contribution; independent evidence is the multi-task empirical results, not an external physical prediction.
  • spatial carrier (output patch as channel-agnostic lattice cell)
    purpose: Conceptual framing that an output patch can hold task quantities instead of RGB appearance.
    Rhetorical device that organizes the argument; not a new physical entity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From RGB Generation to Dense Field Readout: Pixel-Space Dense Prediction with Text-to-Image Models." pith.science (2026). https://pith.science/paper/RXMKZEHT

@misc{pith2026260706553,
  author       = {Pith},
  title        = {Pith review of: From RGB Generation to Dense Field Readout: Pixel-Space Dense Prediction with Text-to-Image Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RXMKZEHT}},
  note         = {Machine review of arXiv:2607.06553}
}
read the original abstract

Large-scale text-to-image models are attractive backbones for dense prediction because RGB generation pretraining learns rich semantic, structural, and geometric priors. Existing generative and editing approaches reuse these priors by casting dense prediction as target generation: annotations such as depth, normals, alpha mattes, masks, and heatmaps are encoded into an RGB-trained VAE latent space and decoded back as image-like targets. We argue this inherits more of the generative output interface than dense prediction requires: unlike RGB synthesis, dense prediction asks for pixel-correct, task-native fields on the same image plane, not new RGB content to be rendered. Our key observation is that a pretrained DiT already organizes RGB inputs through a patch-to-token-to-patch lattice on the image plane, so each token indexes a fixed output patch whose channels can carry task-native quantities instead of RGB appearance. We instantiate this as ReChannel: we keep the VAE encoder for the DiT's input distribution but drop the target-side decoder, adapt the frozen DiT with task LoRA, and map each token to its p x p x K_t pixel-space patch through a shared token-local linear head--about 33K parameters, no spatial mixing. Using FLUX-Klein, we evaluate on six dense prediction tasks and over a dozen benchmarks. This minimal interface sets new state-of-the-art on trimap-free matting, KITTI depth, and referring segmentation, and stays competitive on normals, saliency, and pose. In a matched 4B setting it is more accurate and 2.48x faster than an edit-plus-latent-decode counterpart--dense perception can benefit from generative pretraining without inheriting its output interface.

Figures

Figures reproduced from arXiv: 2607.06553 by the authors.

Figure 1
Figure 1. ReChannel: readout, not generation. A pretrained DiT organizes RGB inputs into a patch-aligned spatial token field, so dense prediction becomes reading out task-native quantities on the same image plane rather than reconstructing an RGB-style target. Each output patch is a spatial carrier in the DiT lattice; the readout reinterprets its channels from RGB appearance to task-native fields—depth, surface normals, matti… view at source ↗
Figure 2
Figure 2. Participation ratio (PR) of the token field: the RGB [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. ReChannel vs. target-side generation. Existing generative and editing interfaces treat dense targets as image-like outputs: the target is encoded into an RGB-trained VAE latent, supervised there, and decoded back through a target-side reconstruction path. ReChannel instead keeps the RGB input interface but drops the target-side path: the pretrained backbone acts as an RGB-native field organizer, and a lightweight li… view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages

  1. [1]

    In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Bae, G., Davison, A.J.: Rethinking inductive bi- ases for surface normal estimation. In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 9535–9545 (2024) 5

    work page 2024

  2. [2]

    co / black - forest - labs / FLUX.2-klein-9B(2026) 2, 5

    Black Forest Labs: FLUX.2 [klein].https : / / huggingface . co / black - forest - labs / FLUX.2-klein-9B(2026) 2, 5

    work page 2026

  3. [3]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Caron, M., Touvron, H., Misra, I., Jégou, H., Mairal, J., Bojanowski, P ., Joulin, A.: Emerging properties in self-supervised vision transformers. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 9650–9660 (2021) 2, 3

    work page 2021

  4. [4]

    In: Proceedings of the 26th ACM international conference on Multi- media

    Chen, Q., Ge, T., Xu, Y., Zhang, Z., Yang, X., Gai, K.: Semantic human matting. In: Proceedings of the 26th ACM international conference on Multi- media. pp. 618–626 (2018) 5

    work page 2018

  5. [5]

    In: Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition

    Cheng, B., Misra, I., Schwing, A.G., Kirillov, A., Girdhar, R.: Masked-attention mask transformer for universal image segmentation. In: Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 1290–1299 (2022) 2, 3

    work page 2022

  6. [6]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weis- senborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al.: An image is worth 16x16 words: Transform- ers for image recognition at scale. arXiv preprint arXiv:2010.11929 (2020) 2, 3

    work page Pith review arXiv 2010

  7. [7]

    In: Pro- ceedings of the IEEE/CVF International Confer- ence on Computer Vision

    Eftekhar, A., Sax, A., Malik, J., Zamir, A.: Om- nidata: A scalable pipeline for making multi-task mid-level vision datasets from 3d scans. In: Pro- ceedings of the IEEE/CVF International Confer- ence on Computer Vision. pp. 10786–10796 (2021) 5

    work page 2021

  8. [8]

    In: European Confer- ence on Computer Vision

    Fu, X., Yin, W., Hu, M., Wang, K., Ma, Y., Tan, P ., Shen, S., Lin, D., Long, X.: Geowizard: Unleash- ing the diffusion priors for 3d geometry estima- tion from a single image. In: European Confer- ence on Computer Vision. pp. 241–258. Springer (2024) 2, 3, 5

    work page 2024

Show all 50 references
  1. [9]

    arXiv preprint arXiv:2604.20329 (2026) 3

    Gabeur, V ., Long, S., Peng, S., Voigtlaender, P ., Sun, S., Bao, Y., Truong, K., Wang, Z., Zhou, W., Barron, J.T., et al.: Image generators are generalist vision learners. arXiv preprint arXiv:2604.20329 (2026) 3

    2026 arXiv

  2. [10]

    In: International Conference on Learning Representations

    He, J., Li, H., Yin, W., Liang, Y., Li, L., Zhou, K., Zhang, H., Liu, B., Chen, Y.: Lotus: Diffusion- based visual foundation model for high-quality dense prediction. In: International Conference on Learning Representations. vol. 2025, pp. 89454– 89467 (2025) 2, 3, 5

    2025

  3. [11]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Ke, B., Obukhov, A., Huang, S., Metzger, N., Daudt, R.C., Schindler, K.: Repurposing diffusion-based image generators for monocu- lar depth estimation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 9492–9502 (2024) 2, 3, 5

    2024

  4. [12]

    In: Proceedings of the AAAI conference on artificial intelligence

    Ke, Z., Sun, J., Li, K., Yan, Q., Lau, R.W.: Modnet: Real-time trimap-free portrait matting via objec- tive decomposition. In: Proceedings of the AAAI conference on artificial intelligence. vol. 36, pp. 1140–1147 (2022) 5

    2022

  5. [13]

    In: Proceedings of the IEEE/CVF international con- ference on computer vision

    Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rol- land, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A.C., Lo, W.Y., et al.: Segment anything. In: Proceedings of the IEEE/CVF international con- ference on computer vision. pp. 4015–4026 (2023) 2, 3

    2023

  6. [14]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Lai, X., Tian, Z., Chen, Y., Li, Y., Yuan, Y., Liu, S., Jia, J.: Lisa: Reasoning segmentation via large language model. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 9579–9589 (2024) 6

    2024

  7. [15]

    In: In- ternational Conference on Learning Representa- tions

    Lan, M., Chen, C., Zhou, Y., Xu, J., Ke, Y., Wang, X., Feng, L., Zhang, W.: Text4seg: Reimagining image segmentation as text generation. In: In- ternational Conference on Learning Representa- tions. vol. 2025, pp. 1634–1661 (2025) 6

    2025

  8. [16]

    In: Proceedings of the 29th ACM international conference on mul- timedia

    Li, J., Ma, S., Zhang, J., Tao, D.: Privacy- preserving portrait matting. In: Proceedings of the 29th ACM international conference on mul- timedia. pp. 3501–3509 (2021) 5

    2021

  9. [17]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, T., He, K.: Back to basics: Let denoising gen- erative models denoise. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 36115–36125 (2026) 3

    2026

  10. [18]

    In: Pro- ceedings of the IEEE/CVF International confer- ence on computer vision

    Li, Y., Zhang, S., Wang, Z., Yang, S., Yang, W., Xia, S.T., Zhou, E.: Tokenpose: Learning key- point tokens for human pose estimation. In: Pro- ceedings of the IEEE/CVF International confer- ence on computer vision. pp. 11313–11322 (2021) 7

    2021

  11. [19]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Liu, J., Ding, H., Cai, Z., Zhang, Y., Satzoda, R.K., Mahadevan, V ., Manmatha, R.: Poly- former: Referring image segmentation as se- quential polygon generation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 18653–18663 (2023) 6

    2023

  12. [20]

    IEEE Transactions on Pattern Analysis and Ma- chine Intelligence46(11), 7300–7316 (2024) 7

    Liu, N., Luo, Z., Zhang, N., Han, J.: Vst++: Ef- ficient and stronger visual saliency transformer. IEEE Transactions on Pattern Analysis and Ma- chine Intelligence46(11), 7300–7316 (2024) 7

    2024

  13. [21]

    IEEE Transactions on Pattern Analy- sis and Machine Intelligence (2025) 7

    Liu, W., Shen, X., Pun, C.M., Cun, X.: Explicit visual prompting for universal foreground seg- mentations. IEEE Transactions on Pattern Analy- sis and Machine Intelligence (2025) 7

    2025

  14. [22]

    In: Pro- ceedings of the IEEE/CVF conference on com- puter vision and pattern recognition

    Lu, P ., Jiang, T., Li, Y., Li, X., Chen, K., Yang, W.: Rtmo: Towards high-performance one-stage real-time multi-person pose estimation. In: Pro- ceedings of the IEEE/CVF conference on com- puter vision and pattern recognition. pp. 1491– 1500 (2024) 7 8

    2024

  15. [23]

    arXiv preprint arXiv:2510.11173 (2025) 6

    Lu, Z., Li, L., Wang, J., Feng, Y., Chen, B., Chen, K., Wang, Y.: Coprs: Learning positional prior from chain-of-thought for reasoning segmenta- tion. arXiv preprint arXiv:2510.11173 (2025) 6

    2025

  16. [24]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Luo, Z., Liu, N., Zhao, W., Yang, X., Zhang, D., Fan, D.P ., Khan, F., Han, J.: Vscode: General visual salient and camouflaged object detection with 2d prompt learning. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 17169–17180 (2024) 7

    2024

  17. [25]

    International journal of computer vision 131(8), 2172–2197 (2023) 5

    Ma, S., Li, J., Zhang, J., Zhang, H., Tao, D.: Rethinking portrait matting with privacy pre- serving. International journal of computer vision 131(8), 2172–2197 (2023) 5

    2023

  18. [26]

    arXiv preprint arXiv:2304.07193 (2023) 2, 3

    Oquab, M., Darcet, T., Moutakanni, T., Vo, H., Szafraniec, M., Khalidov, V ., Fernandez, P ., Haz- iza, D., Massa, F., El-Nouby, A., et al.: Dinov2: Learning robust visual features without supervi- sion. arXiv preprint arXiv:2304.07193 (2023) 2, 3

    2023 arXiv

  19. [27]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Peebles, W., Xie, S.: Scalable diffusion mod- els with transformers. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 4195–4205 (2023) 2, 3

    2023

  20. [28]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Qian, R., Yin, X., Dou, D.: Reasoning to attend: Try to understand how< seg> token works. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 24722–24731 (2025) 6

    2025

  21. [29]

    In: Pro- ceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition

    Qiao, Y., Liu, Y., Yang, X., Zhou, D., Xu, M., Zhang, Q., Wei, X.: Attention-guided hierarchical structure aggregation for image matting. In: Pro- ceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition. pp. 13676– 13685 (2020) 5

    2020

  22. [30]

    In: Proceed- ings of the IEEE/CVF international conference on computer vision

    Ranftl, R., Bochkovskiy, A., Koltun, V .: Vision transformers for dense prediction. In: Proceed- ings of the IEEE/CVF international conference on computer vision. pp. 12179–12188 (2021) 2, 3

    2021

  23. [31]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Rasheed, H., Maaz, M., Shaji, S., Shaker, A., Khan, S., Cholakkal, H., Anwer, R.M., Xing, E., Yang, M.H., Khan, F.S.: Glamm: Pixel grounding large multimodal model. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 13009–13018 (2024) 6

    2024

  24. [32]

    arXiv preprint arXiv:2509.20427 (2025) 2

    Seedream, T., Chen, Y., Gao, Y., Gong, L., Guo, M., Guo, Q., Guo, Z., Hou, X., Huang, W., Huang, Y., et al.: Seedream 4.0: Toward next- generation multimodal image generation. arXiv preprint arXiv:2509.20427 (2025) 2

    2025 arXiv

  25. [33]

    arXiv preprint arXiv:2603.14957 (2026) 3

    Shan, X., Shen, H., Mao, Y., Zhang, X., Anand, A., Li, B., Xu, H., Tu, Z.: Cyclegen: Cycle- consistent layout prediction and image genera- tion in vision foundation models. arXiv preprint arXiv:2603.14957 (2026) 3

    2026

  26. [34]

    arXiv preprint arXiv:2511.18673 (2025) 2, 3, 5

    Shi, Y., Song, Y., Shou, M.Z.: Edit2perceive: Image editing diffusion models are strong dense perceivers. arXiv preprint arXiv:2511.18673 (2025) 2, 3, 5

    2025

  27. [35]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Sun, K., Xiao, B., Liu, D., Wang, J.: Deep high-resolution representation learning for hu- man pose estimation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 5693–5703 (2019) 7

    2019

  28. [36]

    arXiv preprint arXiv:2511.22699 (2025) 2

    Team, Z.I., Cai, H., Cao, S., Du, R., Gao, P ., Hoi, S., Hou, Z., Huang, S., Jiang, D., Jin, X., Li, L., Li, Z., Li, Z.Y., Liu, D., Liu, D., Shi, J., Wu, Q., Yu, F., Zhang, C., Zhang, S., Zhou, S.: Z-image: An efficient image generation foundation model with single-stream diff...

    2025 arXiv

  29. [37]

    arXiv preprint arXiv:2509.04338 (2025) 2, 3

    Wang, J., Lin, C., Sun, L., Liu, R., Nie, L., Li, M., Liao, K., Chu, X.: From editor to dense geometry estimator. arXiv preprint arXiv:2509.04338 (2025) 2, 3

    2025

  30. [38]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Wang, Y., Wang, R., Fan, X., Wang, T., He, X.: Pixels, regions, and objects: Multiple enhance- ment for salient object detection. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 10031–10040 (2023) 7

    2023

  31. [39]

    arXiv preprint arXiv:2510.06139 (2025) 2

    Wang, Z., Jiang, D., Li, L., Dang, S., Li, C., Yang, H., Dai, G., Wang, M., Wang, J.: De- forming videos to masks: Flow matching for referring video segmentation. arXiv preprint arXiv:2510.06139 (2025) 2

    2025

  32. [40]

    In: Proceedings of the IEEE/CVF con- ference on computer vision and pattern recogni- tion

    Wang, Z., Lu, Y., Li, Q., Tao, X., Guo, Y., Gong, M., Liu, T.: Cris: Clip-driven referring image seg- mentation. In: Proceedings of the IEEE/CVF con- ference on computer vision and pattern recogni- tion. pp. 11686–11695 (2022) 6

    2022

  33. [41]

    arXiv preprint arXiv:2508.02324 (2025) 2

    Wu, C., Li, J., Zhou, J., Lin, J., Gao, K., Yan, K., Yin, S.m., Bai, S., Xu, X., Chen, Y., et al.: Qwen-image technical report. arXiv preprint arXiv:2508.02324 (2025) 2

    2025 arXiv

  34. [42]

    Xu, G., Liu, M., Fan, C., Xie, K., Zhao, Z., Chen, H., Shen, C., et al.: What matters when repur- posing diffusion models for general dense per- ception tasks? In: International Conference on Learning Representations. vol. 2025, pp. 6786– 6799 (2025) 2, 3, 5

    2025

  35. [43]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Xu, Y., Zhao, L., Gong, C., Li, G., Wang, D., Wang, N.: Dynpose: Largely improving the efficiency of human pose estimation by a simple dynamic framework. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 1160–1169 (2025) 7

    2025

  36. [44]

    Advances in neural information processing systems35, 38571–38584 (2022) 2, 3, 7

    Xu, Y., Zhang, J., Zhang, Q., Tao, D.: Vitpose: Simple vision transformer baselines for human pose estimation. Advances in neural information processing systems35, 38571–38584 (2022) 2, 3, 7

    2022

  37. [45]

    In: Proceedings of 9 the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Yang, J., Yang, S., Duan, B., Dai, M., Zhang, W., Tan, X., Chen, K., He, W., Wang, J., Wang, H.: Hugging visual prompt and segmentation to- kens: Consistency learning for fine-grained vi- sual understanding in mllms. In: Proceedings of 9 the IEEE/CVF Conference on Computer Vis...

    2026

  38. [46]

    Advances in Neural Information Processing Systems37, 21875–21911 (2024) 2, 3, 5

    Yang, L., Kang, B., Huang, Z., Zhao, Z., Xu, X., Feng, J., Zhao, H.: Depth anything v2. Advances in Neural Information Processing Systems37, 21875–21911 (2024) 2, 3, 5

    2024

  39. [47]

    Advances in neural in- formation processing systems37, 71737–71767 (2024) 6

    Zhang, T., Li, X., Fei, H., Yuan, H., Wu, S., Ji, S., Loy, C.C., Yan, S.: Omg-llava: Bridg- ing image-level, object-level, pixel-level reason- ing and understanding. Advances in neural in- formation processing systems37, 71737–71767 (2024) 6

    2024

  40. [48]

    In: European Conference on Computer Vision

    Zhang, Z., Ma, Y., Zhang, E., Bai, X.: Psalm: Pixelwise segmentation with large multi-modal model. In: European Conference on Computer Vision. pp. 74–91. Springer (2024) 6

    2024

  41. [49]

    arXiv preprint arXiv:2606.15162 (2026) 3

    Zhao, Y., Wang, Y., Wang, X., Wu, Y., Zhang, H., Haji-Ali, M., Abdal, R., Mirzaei, A., Li, Y., Mena- pace, W., et al.: Geostream: Toward precise cam- era controlled streaming video generation. arXiv preprint arXiv:2606.15162 (2026) 3

    2026

  42. [50]

    IEEE Transactions on Pattern Analysis and Machine Intelligence45(3), 3738–3752 (2022) 7 10

    Zhuge, M., Fan, D.P ., Liu, N., Zhang, D., Xu, D., Shao, L.: Salient object detection via integrity learning. IEEE Transactions on Pattern Analysis and Machine Intelligence45(3), 3738–3752 (2022) 7 10

    2022

Pith tools

Reviewed July 11, 2026 · model on record in the stance chip above.