Pith. sign in

REVIEW 3 major objections 5 minor 42 references

Memory Efficient Matting with Adaptive Token Routing

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a content-aware router can send only the informative image tokens through global attention in a matting transformer, cutting peak memory by about 88% and latency by about 50% while keeping matte quality close to the…

desk verdict Memory and latency savings check out and the new dataset is a real contribution, but the router's 'global' feature is constant across tokens and cannot influence routing, so the local-global story does not hold. read the letter →

arxiv 2412.10702 v2 pith:5ZJKQG33 submitted 2024-12-14 cs.CV

Yiheng Lin , Yihan Hu , Chenyi Zhang , Ting Liu , Xiaochao Qu , Luoqi Liu , Yao Zhao , Yunchao Wei This is my paper · ORCID
classification cs.CV
keywords imagemattingadaptivetokenroutingmemory-efficienttransformershigh-resolutionalphamatteknowledgedistillationcompressionUHR-395
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

This paper tries to make transformer-based image matting practical at high resolution by attacking the quadratic memory cost of global self-attention. Its proposed framework, MEMatte, puts a lightweight router before each global attention block: the router scores every image token, sends a content-dependent subset to the global branch, and refreshes the rest through a cheap Lightweight Token Refinement Module instead of dropping them. The paper claims this keeps alpha-matte quality within about 2% on the Composition-1K SAD metric while reducing memory use by roughly 88% and latency by 50%, and it contributes an ultra high-resolution dataset, UHR-395, where full-resolution inference was previously out of reach for most methods.

What carries the argument

The load-bearing mechanism is the combination of the router, BATR, and the Lightweight Token Refinement Module (LTRM) acting inside each transformer block. The router computes a per-token log-probability from a local feature (the first half of a projected token) and a global feature (the average of the second half over all tokens), so each scoring decision has both fine-grained and context information; BATR makes the final 0/1 decision, reselecting every token at every block so no information is permanently discarded, with only a batch-level compression loss steering the average attention ratio; and the LTRM updates the non-routed tokens with depthwise convolution and efficient channel attention so they are refined rather than skipped. An inference-time cap on the number of attended tokens is what ultimately lets ultra high-resolution images fit on a single consumer GPU.

What would settle it

Retrain MEMatte with the learned router replaced by random routing that sends the same average fraction of tokens to global attention; if the sum-of-absolute-differences matte error (SAD) on Composition-1K stays near 21.90, the router's scoring is not what preserves quality, whereas a large SAD jump would confirm the routing decision is load-bearing.

Watch

Extended reading notes

Core claim

The central claim is that adaptive token routing can replace full global self-attention in image matting without a meaningful quality cost. Each router predicts a two-class routing probability from a local token feature concatenated with a globally pooled feature; the Batch-constrained Adaptive Token Routing mechanism (BATR) turns those probabilities into per-block binary decisions with Gumbel-Softmax during training and argmax at inference, and it constrains only the batch-level average attention ratio, so the attended token set adapts to image content and network stage. The non-attended tokens are processed by the Lightweight Token Refinement Module, a depthwise-convolution-plus-channel-attention block taught by distillation to imitate global attention. On the Composition-1K benchmark, MEMatte-S is reported to drop peak memory from 6.20 GB to 0.71 GB and latency from 186 ms to 84.99 ms while raising SAD from 21.46 to 21.90, and on the new UHR-395 and PPT-344 high-resolution tests the full-resolution outputs beat downsampling and crop-and-stitch variants of earlier methods.

Load-bearing premise

The efficiency gain rests on the router's local-global scoring reliably identifying which tokens actually need global attention; if it misroutes informative tokens, matte quality degrades and the memory savings come at the cost of accuracy.

Editorial extensions

If this is right

  • If the reported numbers hold, full-resolution matting of 4K and 8K images becomes feasible on consumer GPUs, eliminating the artifacts of downsampling and crop-and-stitch pipelines.
  • Because the routing ratio is content- and stage-adaptive rather than fixed, the method distributes compute where it is needed instead of paying a flat pruning cost on every image.
  • The ablation indicates that the quality retention depends on distillation and LTRM working together; using either alone does not close the gap to the teacher.
  • The UHR-395 dataset provides a high-resolution benchmark where many prior matting models run out of memory, so future efficiency claims can be tested at the resolution that motivated the method.

Reading between the lines

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

  • Beyond the paper, the same local-global router design could transfer to other dense prediction tasks with full-resolution inputs, such as segmentation, depth estimation, or document analysis, wherever a few informative tokens need global context and the rest can be refined locally.
  • Beyond the paper, the router's scores could serve as a built-in saliency or uncertainty map, since the visualizations indicate that attention is concentrated on object boundaries and fine textures; that could be tested by correlating the scores with boundary annotations.
  • Beyond the paper, the inference-time maximum token count k is a single memory-quality knob, and one could extend MEMatte by predicting k per image instead of setting it globally, lowering memory further on easy images.
  • Beyond the paper, combining adaptive routing with a hierarchical or sparse-attention backbone could compound the savings, since the two mechanisms reduce cost at different scales.
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

3 major / 5 minor

Summary. The manuscript proposes MEMatte, a ViT-based matting framework that inserts a router before each global self-attention block. The router predicts, for every token, whether it should be processed by global attention or by a lightweight token refinement module (LTRM); a batch-constrained mechanism adapts the routed fraction rather than using a fixed ratio, and a distillation loss transfers knowledge from a frozen ViTMatte teacher. The paper also introduces UHR-395, a large ultra-high-resolution matting dataset. On Composition-1K, MEMatte-S/B report roughly 88% memory reduction and about 50% latency reduction relative to ViTMatte, with small SAD increases, and it reports improved results on high-resolution and real-world benchmarks.

Significance. If the reported results hold, the efficiency contribution is practically valuable: full-resolution matting at 4K/8K on a single consumer GPU is a real enabler, and the dataset is a plausible resource for the community. The paper also contains useful ablations (routing compression, maximum token cap, distillation/LTRM) and release code. However, the paper's core architectural novelty—local-global routing—is undermined by a mathematical issue in the router definition, and the high-resolution comparisons mix the efficiency contribution with an input-resolution confound. The measured savings are still consistent with the architecture, but the paper needs revision before it can be accepted.

major comments (3)
  1. [Adaptive Token Routing, Eqs. (5)-(7)] The global feature zg_i defined in Eq. (5) is the same vector for every token i. In Eq. (7), f'_theta is a linear layer followed by LogSoftmax, so the contribution W_g zg_i is an additive constant across all tokens and cancels in the softmax/argmax. Consequently the routing probability p_i and decision delta_i depend only on the local feature zl_i; the 'local-global strategy' is formally present but decision-inert. This removes the stated mechanism by which the router identifies globally informative tokens and weakens the interpretation of Figure 5, which is equally consistent with a purely local edge/texture heuristic. Please either modify the router so that the global feature can affect per-token decisions (for example by using a non-constant interaction with local features) or rewrite the methodology and claims to describe a local-only router.
  2. [Tables 3 and 4] The high-resolution comparisons let ViTMatte run only on downsampled input (D) or patches (P), while MEMatte runs on full-resolution input. This confounds the efficiency of the routing architecture with the resolution of the input. The reported gains on UHR-395 and PPT-344 could be driven substantially by the full-resolution processing rather than by adaptive token routing. Please add a full-resolution ViTMatte baseline on PPT-344 (which is small enough to be feasible) or otherwise quantify the contribution of input resolution, and adjust the 'outperforms existing methods' wording accordingly.
  3. [Adaptation to Ultra High-Resolution Images and Figure 6] The maximum token cap k is introduced and ablated, but the default value used for the reported results is never given in the main text or in Table 2 or Table 7. Please report the default k and the exact inference recipe (per-stage or global cap). In addition, all quantitative tables report a single run without error bars or seeds, which is insufficient to support the fine performance differences claimed, such as the 0.02-0.11 MSE differences on AIM-500 in Table 5.
minor comments (5)
  1. [Eq. (5)] The notation in Eq. (5) uses the subscript i on the left-hand side while the right-hand side averages over all tokens; please use a pooled symbol such as z^g without the token index to avoid implying a per-token global feature.
  2. [Table 2] The formatted values '0.7188.5%↓' and '84.9954.3%↓' are difficult to read; please insert spaces or use separate columns for memory/latency and the percentage reduction.
  3. [Abstract and Experiments] The abstract states the average resolution of UHR-395 is 4872×6017, while the experiments section states the test set average is 5318×7051; please clarify which number applies to the whole dataset and which to the test subset.
  4. [Figure 2] The figure caption mentions the RTX 1060 in the text but the experiments are run on the RTX 3090; please ensure all GPUs used for the memory curves are identified in the figure or caption.
  5. [Table 3] MEMatte-S* is fine-tuned on UHR-395, but the ViTMatte baselines in the same table are not fine-tuned on UHR-395; please add a fine-tuned ViTMatte baseline or state why such a comparison is not possible.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: benchmark metrics are measured against a frozen external ViTMatte teacher, and the router-global-feature concern is an architectural nuance, not a circular derivation.

full rationale

I walked the claimed derivation chain. The router (Eqs. 3-7) outputs routing probabilities, BATR (Eq. 8) makes discrete decisions via Gumbel-Softmax or argmax, LTRM (Eq. 10) is a lightweight refinement branch, and the losses (Eqs. 11-13) use a frozen external ViTMatte teacher plus a compression regularizer. None of these equations defines a headline benchmark number in terms of the value being predicted: memory, latency, SAD, MSE, Grad, and Conn in Tables 2-5 are measured after training, not algebraically derived from rho, k, or the router scores. The distillation setup compares MEMatte with the same ViTMatte model used as teacher, but the teacher is a fixed external checkpoint and the student's metrics are evaluated on held-out sets, so this is standard practice rather than a logical circle. The only self-referential flavor is that several related-work citations share authors with the present paper, but none of these is load-bearing for the memory-efficiency or routing claims. I also considered the Eq. 5 concern that the global feature is an average over all tokens and hence token-invariant; even taken at face value, this is an architectural limitation about whether global context can re-rank tokens, not a circular derivation in which a result is equivalent to its own inputs. Therefore the circularity score is 0.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The ledger shows the method rests on two tuned hyperparameters (rho and k), on standard attention complexity math, and on several domain assumptions about routing informativeness, the sufficiency of the lightweight branch, and the validity of synthetic high-res composites. There are no invented physical entities; UHR-395 is a dataset resource rather than a postulated entity.

free parameters (2)
  • target compression degree rho = 0.25 (default)
    Chosen in Table 7 from an ablation over 0.75, 0.50, 0.25, and 0.10 as the best memory-performance trade-off; Eq. 12 uses it to constrain the batch-level attention ratio gamma.
  • maximum token cap k = not stated in main text
    Introduced in the 'Adaptation to Ultra High-Resolution Images' section to bound attention tokens at inference; Figure 6 shows sensitivity but the default value is only in the supplementary, and memory and latency numbers depend on it.
assumptions (6)
  • standard math Global self-attention has O(N^2) memory complexity as defined in Eq. 2.
    Used throughout to justify the router and LTRM design; it is a textbook property of the attention operation.
  • domain assumption In early ViT stages, global attention behaves approximately like convolution, so few tokens need full attention there.
    Invoked in the introduction and in the Visualization of Token Routing discussion, citing Park and Kim 2022 and Ghiasi et al. 2022; if false, routing few tokens in early blocks could lose useful context.
  • domain assumption A lightweight branch made of depthwise convolution and ECA can refine uninformative tokens as well as global attention when trained with distillation.
    This is the load-bearing premise of the LTRM in Eq. 10; the paper supports it with the combined ablation in Table 8 but not with direct feature-level equivalence.
  • ad hoc to paper The local-global router, whose global feature is a single average over all tokens (Eq. 5), carries enough context to identify informative tokens.
    The averaging of the second half of channels over all tokens is a design choice; no analysis is provided on how much global context it preserves.
  • domain assumption Ultra-high-resolution synthetic composites in UHR-395 are a valid benchmark for high-resolution matting quality.
    The dataset is constructed by compositing 395 alpha mattes onto backgrounds under the DIM rule; the paper argues this is needed because existing datasets lack resolution and diversity, but no validation is given that composite difficulty matches real high-res matting.
  • domain assumption Models trained on DIM transfer to UHR-395, PPT-344, and AIM-500 without additional adaptation (except the fine-tuned starred variant).
    All main comparisons follow the DIM training protocol and evaluate on other datasets, which is standard in matting but is an assumption about distribution shift.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Memory Efficient Matting with Adaptive Token Routing." pith.science (2026). https://pith.science/paper/5ZJKQG33

@misc{pith2026241210702,
  author       = {Pith},
  title        = {Pith review of: Memory Efficient Matting with Adaptive Token Routing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5ZJKQG33}},
  note         = {Machine review of arXiv:2412.10702}
}
abstract

Transformer-based models have recently achieved outstanding performance in image matting. However, their application to high-resolution images remains challenging due to the quadratic complexity of global self-attention. To address this issue, we propose MEMatte, a \textbf{m}emory-\textbf{e}fficient \textbf{m}atting framework for processing high-resolution images. MEMatte incorporates a router before each global attention block, directing informative tokens to the global attention while routing other tokens to a Lightweight Token Refinement Module (LTRM). Specifically, the router employs a local-global strategy to predict the routing probability of each token, and the LTRM utilizes efficient modules to simulate global attention. Additionally, we introduce a Batch-constrained Adaptive Token Routing (BATR) mechanism, which allows each router to dynamically route tokens based on image content and the stages of attention block in the network. Furthermore, we construct an ultra high-resolution image matting dataset, UHR-395, comprising 35,500 training images and 1,000 test images, with an average resolution of $4872\times6017$. This dataset is created by compositing 395 different alpha mattes across 11 categories onto various backgrounds, all with high-quality manual annotation. Extensive experiments demonstrate that MEMatte outperforms existing methods on both high-resolution and real-world datasets, significantly reducing memory usage by approximately 88% and latency by 50% on the Composition-1K benchmark. Our code is available at https://github.com/linyiheng123/MEMatte.

Figures

Figures reproduced from arXiv: 2412.10702 by the authors.

Figure 1
Figure 1. Illustration of the importance of full-resolution in [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Memory Usage / Image Resolution. OOM de [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overall framework of the proposed MEMatte. The router module is inserted before global attention to predict the [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison of the results on the UHR-395 test set. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Visualization of the token routing. The retained tokens are routed to global attention branch, while the gray tokens are [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Effect of reducing maximum token number k. Effect of Target Compression degree [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 27 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Bolya, D.; Fu, C.-Y.; Dai, X.; Zhang, P.; Feichtenhofer, C.; and Hoffman, J. 2023. Token Merging: Your ViT but Faster. In International Conference on Learning Representations

    work page 2023

  4. [4]

    Cai, H.; Xue, F.; Xu, L.; and Guo, L. 2022. Transmatting: Enhancing transparent objects matting with transformers. In European conference on computer vision, 253--269. Springer

    work page 2022

  5. [5]

    Chen, Q.; Li, D.; and Tang, C.-K. 2013. KNN matting. IEEE transactions on pattern analysis and machine intelligence, 35(9): 2175--2188

    work page 2013

  6. [6]

    Dai, Y.; Price, B.; Zhang, H.; and Shen, C. 2022. Boosting robustness of image matting with context assembling and strong data augmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11707--11716

    work page 2022

  7. [7]

    Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

    arXiv 2020

  8. [8]

    Z.; Tong, C

    Engelcke, M.; Rao, D.; Wang, D. Z.; Tong, C. H.; and Posner, I. 2017. Vote3deep: Fast object detection in 3d point clouds using efficient convolutional neural networks. In 2017 IEEE International Conference on Robotics and Automation (ICRA), 1355--1361. IEEE

    work page 2017

Show all 42 references
  1. [9]

    K.; Winn, J.; and Zisserman, A

    Everingham, M.; Van Gool, L.; Williams, C. K.; Winn, J.; and Zisserman, A. 2010. The pascal visual object classes (voc) challenge. International journal of computer vision, 88: 303--338

    2010

  2. [10]

    Forte, M.; and Piti \'e , F. 2020. F , B , Alpha Matting. arXiv preprint arXiv:2003.07711

    2020 arXiv

  3. [11]

    S.; and Oliveira, M

    Gastal, E. S.; and Oliveira, M. M. 2010. Shared sampling for real-time alpha matting. In Computer Graphics Forum, volume 29, 575--584. Wiley Online Library

    2010

  4. [12]

    G.; and Goldstein, T

    Ghiasi, A.; Kazemi, H.; Borgnia, E.; Reich, S.; Shu, M.; Goldblum, M.; Wilson, A. G.; and Goldstein, T. 2022. What do vision transformers learn? a visual exploration. arXiv preprint arXiv:2212.06727

    2022 arXiv

  5. [13]

    He, K.; Rhemann, C.; Rother, C.; Tang, X.; and Sun, J. 2011. A global sampling method for alpha matting. In CVPR 2011, 2049--2056. Ieee

    2011

  6. [14]

    He, K.; Sun, J.; and Tang, X. 2010. Fast matting using large kernel matting laplacian matrices. In 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2165--2172. IEEE

    2010

  7. [15]

    Hu, Y.; Lin, Y.; Wang, W.; Zhao, Y.; Wei, Y.; and Shi, H. 2025. Diffusion for natural image matting. In European Conference on Computer Vision, 181--199. Springer

    2025

  8. [16]

    Ke, Z.; Sun, J.; Li, K.; Yan, Q.; and Lau, R. W. 2022. Modnet: Real-time trimap-free portrait matting via objective decomposition. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, 1140--1147

    2022

  9. [17]

    Levin, A.; Lischinski, D.; and Weiss, Y. 2007. A closed-form solution to natural image matting. IEEE transactions on pattern analysis and machine intelligence, 30(2): 228--242

    2007

  10. [18]

    Li, J.; Goel, V.; Ohanyan, M.; Navasardyan, S.; Wei, Y.; and Shi, H. 2024 a . Vmformer: End-to-end video matting with transformer. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 6678--6687

    2024

  11. [19]

    Li, J.; Zhang, J.; and Tao, D. 2021. Deep automatic natural image matting. arXiv preprint arXiv:2107.07235

    2021 arXiv

  12. [20]

    Li, Y.; Huang, Z.; Yu, G.; Chen, L.; Wei, Y.; and Jiao, J. 2024 b . Disentangled Pre-training for Image Matting. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 169--178

    2024

  13. [21]

    Li, Y.; and Lu, H. 2020. Natural image matting via guided contextual attention. In Proceedings of the AAAI conference on artificial intelligence, volume 34, 11450--11457

    2020

  14. [22]

    Li, Y.; Mao, H.; Girshick, R.; and He, K. 2022. Exploring plain vision transformer backbones for object detection. In European conference on computer vision, 280--296. Springer

    2022

  15. [23]

    Liang, Y.; Ge, C.; Tong, Z.; Song, Y.; Wang, J.; and Xie, P. 2022. Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations. In International Conference on Learning Representations

    2022

  16. [24]

    L.; Seitz, S

    Lin, S.; Ryabtsev, A.; Sengupta, S.; Curless, B. L.; Seitz, S. M.; and Kemelmacher-Shlizerman, I. 2021. Real-time high-resolution background matting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 8762--8771

    2021

  17. [25]

    Lin, T.-Y.; Maire, M.; Belongie, S.; Hays, J.; Perona, P.; Ramanan, D.; Doll \'a r, P.; and Zitnick, C. L. 2014. Microsoft coco: Common objects in context. In Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 1...

    2014

  18. [26]

    Liu, H.; Tan, Z.; Tan, C.; Wei, Y.; Wang, J.; and Zhao, Y. 2024. Forgery-aware adaptive transformer for generalizable synthetic image detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10770--10780

    2024

  19. [27]

    Liu, Z.; Lin, Y.; Cao, Y.; Hu, H.; Wei, Y.; Zhang, Z.; Lin, S.; and Guo, B. 2021. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, 10012--10022

    2021

  20. [28]

    Lu, H.; Dai, Y.; Shen, C.; and Xu, S. 2019. Indices Matter: Learning to Index for Deep Image Matting. In Proc. IEEE/CVF International Conference on Computer Vision (ICCV)

    2019

  21. [29]

    Park, G.; Son, S.; Yoo, J.; Kim, S.; and Kwak, N. 2022. Matteformer: Transformer-based image matting via prior-tokens. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11696--11706

    2022

  22. [30]

    Park, N.; and Kim, S. 2022. HOW DO VISION TRANSFORMERS WORK? In 10th International Conference on Learning Representations, ICLR 2022

    2022

  23. [31]

    Qiao, Y.; Liu, Y.; Yang, X.; Zhou, D.; Xu, M.; Zhang, Q.; and Wei, X. 2020. Attention-guided hierarchical structure aggregation for image matting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 13676--13685

    2020

  24. [32]

    Rao, Y.; Zhao, W.; Liu, B.; Lu, J.; Zhou, J.; and Hsieh, C.-J. 2021. Dynamicvit: Efficient vision transformers with dynamic token sparsification. Advances in neural information processing systems, 34: 13937--13949

    2021

  25. [33]

    Sun, Y.; Tang, C.-K.; and Tai, Y.-W. 2021. Semantic image matting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11120--11129

    2021

  26. [34]

    Sun, Y.; Tang, C.-K.; and Tai, Y.-W. 2023. Ultrahigh resolution image/video matting with spatio-temporal sparsity. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 14112--14121

    2023

  27. [35]

    N.; Kaiser, .; and Polosukhin, I

    Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, .; and Polosukhin, I. 2017. Attention is all you need. Advances in neural information processing systems, 30

    2017

  28. [36]

    M.; Porikli, F.; and Habibian, A

    Venkataramanan, S.; Ghodrati, A.; Asano, Y. M.; Porikli, F.; and Habibian, A. 2023. Skip-attention: Improving vision transformers by paying less attention. arXiv preprint arXiv:2301.02240

    2023 arXiv

  29. [37]

    Wang, Q.; Wu, B.; Zhu, P.; Li, P.; Zuo, W.; and Hu, Q. 2020. ECA-Net: Efficient channel attention for deep convolutional neural networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 11534--11542

    2020

  30. [38]

    Xu, N.; Price, B.; Cohen, S.; and Huang, T. 2017. Deep image matting. In Proceedings of the IEEE conference on computer vision and pattern recognition, 2970--2979

    2017

  31. [39]

    Yao, J.; Wang, X.; Yang, S.; and Wang, B. 2024. Vitmatte: Boosting image matting with pre-trained plain vision transformers. Information Fusion, 103: 102091

    2024

  32. [40]

    Yu, H.; Xu, N.; Huang, Z.; Zhou, Y.; and Shi, H. 2021 a . High-resolution deep image matting. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, 3217--3224

    2021

  33. [41]

    Yu, Q.; Zhang, J.; Zhang, H.; Wang, Y.; Lin, Z.; Xu, N.; Bai, Y.; and Yuille, A. 2021 b . Mask guided matting via progressive refinement network. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 1154--1163

    2021

  34. [42]

    Zhang, Y.; Wei, Y.; Jiang, D.; Zhang, X.; Zuo, W.; and Tian, Q. 2023. ControlVideo: Training-free Controllable Text-to-Video Generation. arXiv preprint arXiv:2305.13077

    2023 arXiv

Pith tools

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