pith. sign in

arxiv: 2606.29400 · v1 · pith:BGQZBZ3Xnew · submitted 2026-06-28 · 💻 cs.CV · cs.AI· cs.GR

Learning to Adaptively Allocate Gaussians for Arbitrary-Scale Image Super-Resolution

Giulio Federico , Giuseppe Amato , Claudio Gennaro , Fabio Carrara , Marco Di Benedetto This is my paper

Pith reviewed 2026-06-30 08:05 UTC · model grok-4.3

classification 💻 cs.CV cs.AIcs.GR
keywords arbitrary-scale super-resolutiongaussian splattingfeed-forward networkneural routingadaptive densificationhierarchical pointer convolutionimage upscaling
0
0 comments X

The pith

A feed-forward network learns to predict Gaussian densification for arbitrary-scale super-resolution from low-resolution inputs alone.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces QuADA-GS to convert Gaussian Splatting into a feed-forward model for arbitrary-scale image super-resolution. Standard GS optimization relies on high-resolution gradients unavailable at inference, so the network must instead learn to predict how to densify primitives directly from low-resolution data. It encodes the input, routes computation by assessing local complexity to allocate a global upsampling budget, densifies features into irregular topologies, and decodes them into 2D Gaussians using a new convolution operator for efficient neighbor lookup. This addresses the inefficiency of uniform pixel processing and the limitations of fixed-scale models that fall back on interpolation for continuous resolutions.

Core claim

After encoding low-resolution inputs into latent space, a Neural Routing Architecture evaluates local complexity to distribute a global budget and assign specific upsampling factors to features. Features are then dynamically densified according to these factors to create an irregular topology that is decoded into 2D Gaussian primitives. Hierarchical Pointer Convolution coordinates the features before decoding with O(1) neighbor lookup complexity.

What carries the argument

Neural Routing Architecture that evaluates local complexity to assign upsampling factors and control adaptive densification of features into Gaussian primitives.

If this is right

  • QuADA-GS achieves state-of-the-art performance on arbitrary-scale super-resolution benchmarks.
  • The method maintains low latency while using a lean memory footprint by concentrating resources on complex regions.
  • It supports continuous scaling factors without relying on sub-optimal post-hoc interpolation.
  • Hierarchical Pointer Convolution enables efficient spatial communication without dense grid bottlenecks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The routing mechanism suggests a general way to replace gradient-driven primitive growth with learned complexity prediction in other splatting-based representations.
  • This could reduce memory overhead in real-time graphics pipelines that must handle variable zoom levels or foveated rendering.
  • Extending the same routing logic to 3D or video data might allow adaptive allocation across time or depth without retraining the core densification logic.

Load-bearing premise

A feed-forward network can autonomously predict the correct densification of Gaussian primitives from low-resolution inputs alone, without high-resolution gradients that standard Gaussian Splatting optimization requires during training.

What would settle it

A test showing that QuADA-GS quality collapses or falls below interpolation-based baselines on non-integer scales when the routing network is replaced by uniform allocation would falsify the claim.

Figures

Figures reproduced from arXiv: 2606.29400 by Claudio Gennaro, Fabio Carrara, Giulio Federico, Giuseppe Amato, Marco Di Benedetto.

Figure 1
Figure 1. Figure 1: Differentiable Neural Routing Architecture. The lightweight router R𝜃 processes refined latent features E (𝑟𝑒 𝑓 𝑖𝑛𝑒𝑑) 𝑔𝑠 to predict continuous probability distributions over three upsampling levels. By applying the Gumbel-Softmax trick, these collapse into a discrete routing map M𝑟𝑜𝑢𝑡𝑒 while maintaining differentiability. To prevent over-allocation, the map’s total spatial cost C𝑟𝑜𝑢𝑡𝑒𝑟 is compared to a tar… view at source ↗
Figure 2
Figure 2. Figure 2: Hierarchical Generation Pipeline. For visual clarity, we illustrate the process using a 2 × 2 input E (𝑟𝑒 𝑓 𝑖𝑛𝑒𝑑) 𝑔𝑠 with heterogeneous upsampling factors dictated by the Routing Map. Each phase executes a ×2 split followed by topological coordination via the HPC operator. Hierarchical Pointer Convolution. While cascaded splitting avoids dense computation, independent expansion strips features of their top… view at source ↗
Figure 4
Figure 4. Figure 4: Adaptive Boundary-Hugging Lookup and Feature Update. Il￾lustration of the lookup mechanism retrieving the topological neighborhood N (𝑓 ) for a target feature. By sampling the Pointer Grid at the footprint boundaries, the operator captures adjacent identifiers across varying levels, which are subsequently gathered from the flattened tensor to perform the residual sparse update. Gaussian Parameter Predictio… view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparisons at ×12 between QuADA-GS (and its routing map) and competitors, using RDN [Zhang et al. 2018c] as image encoder [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative comparisons at ×4 between QuADA-GS (and its routing map) and competitors, using RDN [Zhang et al. 2018c] as image encoder [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative comparisons at ×6 between QuADA-GS (and its routing map) and competitors, using RDN [Zhang et al. 2018c] as image encoder [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative comparisons at ×8 between QuADA-GS (and its routing map) and competitors, using RDN [Zhang et al. 2018c] as image encoder [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Qualitative comparisons at ×12 between QuADA-GS (and its routing map) and competitors, using RDN [Zhang et al. 2018c] as image encoder [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Qualitative comparisons at ×16 between QuADA-GS (and its routing map) and competitors, using RDN [Zhang et al. 2018c] as image encoder [PITH_FULL_IMAGE:figures/full_fig_p011_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Qualitative comparisons at ×6 (left) and ×12 (right) evaluating the impact of the backbone architecture on QuADA-GS. We compare QuADA-GS utilizing RDN [Zhang et al. 2018c] against the EDSR_baseline [Lim et al. 2017]. As demonstrated, employing RDN allows the network to extract richer high-frequency features, successfully recovering finer structural details compared to the EDSR baseline [PITH_FULL_IMAGE:f… view at source ↗
read the original abstract

In computer graphics, visual content is continuously warped, zoomed and resampled. This occurs when engines upscale frames, users zoom into 3D scenes, or foveated VR applies varying scaling. Handling these transformations requires Arbitrary-Scale Super-Resolution (ASR). Traditional models, designed for fixed scales, typically predict at a lower integer scale (e.g., x4) and rely on sub-optimal interpolation for continuous resolutions, compromising quality. Furthermore, most methods process pixels uniformly. Since fine details are sparse, this creates overhead; efficiency dictates concentrating resources only where structural complexity demands it. While implicit models and Gaussian Splatting (GS) enable continuous representation, GS is advantageous due to adaptive densification. However, transitioning GS into a feed-forward model for ASR is non-trivial. Standard GS optimization needs high-resolution gradients to drive primitive growth, which are unavailable during inference. Thus, the network must autonomously predict GS densification from low-resolution inputs. To solve this, we propose QuADA-GS. After encoding inputs into a latent space, a Neural Routing Architecture evaluates local complexity to distribute a global budget, assigning specific upsampling factors to features to avoid redundant processing. Features are dynamically densified based on these factors, forming an irregular topology decoded into 2D Gaussian primitives. To coordinate features before decoding, we introduce Hierarchical Pointer Convolution. This non-grid operator achieves O(1) neighbor lookup complexity, facilitating efficient spatial communication and bypassing dense bottlenecks. Experiments show QuADA-GS achieves state-of-the-art ASR performance, maintaining low latency and a lean memory footprint.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript proposes QuADA-GS, a feed-forward architecture for arbitrary-scale image super-resolution. It encodes low-resolution inputs into latent space, employs a Neural Routing Architecture to evaluate local complexity and distribute a global budget by predicting per-feature upsampling factors, dynamically densifies features into an irregular topology of 2D Gaussian primitives, and decodes them via the introduced Hierarchical Pointer Convolution operator (claimed O(1) neighbor lookup) to enable efficient spatial communication. The central claim is that this learned adaptive allocation achieves state-of-the-art ASR performance while maintaining low latency and a lean memory footprint.

Significance. If the empirical results hold and the routing network successfully learns to allocate primitives where structural complexity requires them, the work would be significant for efficient continuous-scale super-resolution. It combines the adaptive densification property of Gaussian Splatting with feed-forward inference, addressing uniform pixel processing overhead in traditional ASR methods and enabling resource concentration on complex regions without dense bottlenecks.

major comments (1)
  1. [Abstract] Abstract: The SOTA performance claim rests on the Neural Routing Architecture autonomously predicting correct densification of Gaussian primitives from LR inputs alone. However, standard GS adaptive densification is driven by HR gradients during optimization; the feed-forward path has no such signal at inference. No quantitative results, ablation studies on routing accuracy, or implementation details are supplied to verify that the learned policy allocates primitives where structural complexity (e.g., edges or textures visible only at target scale) actually demands them, which is load-bearing for both quality and the claimed memory/latency savings.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for highlighting this critical aspect of our contribution. The concern about validating the Neural Routing Architecture's ability to predict appropriate Gaussian densification from LR inputs alone is well-taken, as it underpins both the quality and efficiency claims. We address this point below and will revise the manuscript to incorporate additional supporting evidence.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The SOTA performance claim rests on the Neural Routing Architecture autonomously predicting correct densification of Gaussian primitives from LR inputs alone. However, standard GS adaptive densification is driven by HR gradients during optimization; the feed-forward path has no such signal at inference. No quantitative results, ablation studies on routing accuracy, or implementation details are supplied to verify that the learned policy allocates primitives where structural complexity (e.g., edges or textures visible only at target scale) actually demands them, which is load-bearing for both quality and the claimed memory/latency savings.

    Authors: We agree that direct validation of the routing policy is essential and was insufficiently emphasized. While the end-to-end training on LR-to-HR pairs allows the network to learn a policy that correlates with structural complexity (as evidenced by the reported SOTA quality and efficiency gains over uniform baselines), we acknowledge the absence of explicit routing-accuracy metrics. In the revised version we will add: (1) quantitative comparison of predicted per-feature upsampling factors against proxy ground-truth complexity maps computed from HR gradients on held-out data; (2) ablation studies that disable the Neural Routing Architecture (replacing it with uniform allocation) and report resulting PSNR/SSIM drops as well as changes in memory footprint and latency; and (3) implementation details on the routing loss and budget-distribution mechanism. These additions will directly address the load-bearing claim. revision: yes

Circularity Check

0 steps flagged

No circularity; feed-forward prediction of densification is an empirical claim, not a definitional reduction.

full rationale

The provided abstract and description contain no equations, fitted parameters renamed as predictions, or self-citation chains that reduce the claimed ASR performance or Gaussian allocation to quantities defined by the inputs themselves. The method is presented as a learned architecture (Neural Routing + Hierarchical Pointer Convolution) that must solve the non-trivial mapping from LR to adaptive GS primitives; this is an independent empirical claim supported by experiments rather than a self-referential construction. No load-bearing step matches any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

Abstract-only review prevents exhaustive enumeration; the central claim rests on the unverified assumption that densification can be predicted from low-resolution inputs and on the existence of a global budget hyperparameter.

free parameters (1)
  • global budget
    A global budget for distributing upsampling factors is mentioned; its value is chosen during training and affects the final allocation.
axioms (1)
  • domain assumption Feed-forward networks can learn to replicate the densification behavior of gradient-driven Gaussian Splatting optimization
    The paper states that standard GS optimization needs high-resolution gradients unavailable at inference, so the network must predict densification instead.
invented entities (1)
  • Hierarchical Pointer Convolution no independent evidence
    purpose: Non-grid operator for O(1) neighbor lookup on irregular feature topologies
    New operator introduced to enable efficient spatial communication without dense bottlenecks.

pith-pipeline@v0.9.1-grok · 5825 in / 1219 out tokens · 46095 ms · 2026-06-30T08:05:47.221208+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

74 extracted references · 8 canonical work pages · 3 internal anchors

  1. [1]

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

    Human guided ground-truth generation for realistic image super-resolution , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  2. [2]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Toward generalized image quality assessment: Relaxing the perfect reference quality assumption , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  3. [3]

    Proceedings of the 32nd ACM International Conference on Multimedia , pages=

    SSL: A self-similarity loss for improving generative image super-resolution , author=. Proceedings of the 32nd ACM International Conference on Multimedia , pages=

  4. [4]

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

    Details or artifacts: A locally discriminative learning approach to realistic image super-resolution , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  5. [5]

    International Journal of Computer Vision , volume=

    Exploiting diffusion prior for real-world image super-resolution , author=. International Journal of Computer Vision , volume=. 2024 , publisher=

    2024

  6. [6]

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

    Real-esrgan: Training real-world blind super-resolution with pure synthetic data , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  7. [7]

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

    Sinsr: diffusion-based image super-resolution in a single step , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  8. [8]

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

    Seesr: Towards semantics-aware real-world image super-resolution , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  9. [9]

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

    Fine-structure preserved real-world image super-resolution via transfer vae training , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  10. [10]

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

    Scaling up to excellence: Practicing model scaling for photo-realistic image restoration in the wild , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  11. [11]

    Advances in neural information processing systems , volume=

    Resshift: Efficient diffusion model for image super-resolution by residual shifting , author=. Advances in neural information processing systems , volume=

  12. [12]

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

    Zhang, Kai and Liang, Jingyun and Van Gool, Luc and Timofte, Radu , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2021 , pages =

    2021

  13. [13]

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

    Activating more pixels in image super-resolution transformer , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  14. [14]

    European conference on computer vision , pages=

    Accelerating the super-resolution convolutional neural network , author=. European conference on computer vision , pages=. 2016 , organization=

    2016

  15. [15]

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

    Li, Zhen and Yang, Jinglei and Liu, Zheng and Yang, Xiaomin and Jeon, Gwanggil and Wu, Wei , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

  16. [16]

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

    Liang, Jingyun and Cao, Jiezhang and Sun, Guolei and Zhang, Kai and Van Gool, Luc and Timofte, Radu , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops , month =. 2021 , pages =

    2021

  17. [17]

    Proceedings of the European conference on computer vision (ECCV) workshops , pages=

    Esrgan: Enhanced super-resolution generative adversarial networks , author=. Proceedings of the European conference on computer vision (ECCV) workshops , pages=

  18. [18]

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

    Transcending the limit of local window: Advanced super-resolution transformer with adaptive token dictionary , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  19. [19]

    European conference on computer vision , pages=

    Efficient long-range attention network for image super-resolution , author=. European conference on computer vision , pages=. 2022 , organization=

    2022

  20. [20]

    Proceedings of the European conference on computer vision (ECCV) , pages=

    Image super-resolution using very deep residual channel attention networks , author=. Proceedings of the European conference on computer vision (ECCV) , pages=

  21. [21]

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

    Residual dense network for image super-resolution , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  22. [22]

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

    Learning continuous image representation with local implicit image function , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  23. [23]

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

    Local texture estimator for implicit representation function , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  24. [24]

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

    Yao, Jie-En and Tsao, Li-Yuan and Lo, Yi-Chen and Tseng, Roy and Chang, Chia-Che and Lee, Chun-Yi , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2023 , pages =

    2023

  25. [25]

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

    Ciaosr: Continuous implicit attention-in-attention network for arbitrary-scale image super-resolution , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  26. [26]

    Communications of the ACM , volume=

    Nerf: Representing scenes as neural radiance fields for view synthesis , author=. Communications of the ACM , volume=. 2021 , publisher=

    2021

  27. [27]

    European conference on computer vision , pages=

    Learning a deep convolutional network for image super-resolution , author=. European conference on computer vision , pages=. 2014 , organization=

    2014

  28. [28]

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

    Kim, Jiwon and Lee, Jung Kwon and Lee, Kyoung Mu , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , month =

  29. [29]

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

    Deeply-recursive convolutional network for image super-resolution , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  30. [30]

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

    Deep laplacian pyramid networks for fast and accurate super-resolution , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  31. [31]

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

    Meta-SR: A magnification-arbitrary network for super-resolution , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  32. [32]

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

    Learning a single network for scale-arbitrary super-resolution , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  33. [33]

    Implicit Transformer Network for Screen Content Image Continuous Super-Resolution , url =

    Yang, Jingyu and Shen, Sheng and Yue, Huanjing and Li, Kun , booktitle =. Implicit Transformer Network for Screen Content Image Continuous Super-Resolution , url =

  34. [34]

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

    Wei, Min and Zhang, Xuesong , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2023 , pages =

    2023

  35. [35]

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

    Chen, Hao-Wei and Xu, Yu-Syuan and Hong, Min-Fong and Tsai, Yi-Min and Kuo, Hsien-Kai and Lee, Chun-Yi , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2023 , pages =

    2023

  36. [36]

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

    He, Zongyao and Jin, Zhi , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2024 , pages =

    2024

  37. [37]

    , author=

    3d gaussian splatting for real-time radiance field rendering. , author=. ACM Trans. Graph. , volume=

  38. [38]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Gaussiansr: High fidelity 2d gaussian splatting for arbitrary-scale image super-resolution , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  39. [39]

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

    Chen, Du and Chen, Liyi and Zhang, Zhengqiang and Zhang, Lei , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2025 , pages =

    2025

  40. [40]

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

    Lim, Bee and Son, Sanghyun and Kim, Heewon and Nah, Seungjun and Mu Lee, Kyoung , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops , month =

  41. [41]

    Proceedings of the European Conference on Computer Vision (ECCV) , month =

    Zhang, Yulun and Li, Kunpeng and Li, Kai and Wang, Lichen and Zhong, Bineng and Fu, Yun , title =. Proceedings of the European Conference on Computer Vision (ECCV) , month =

  42. [42]

    Categorical Reparameterization with Gumbel-Softmax

    Categorical reparameterization with gumbel-softmax , author=. arXiv preprint arXiv:1611.01144 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  43. [43]

    and Bishop, Rob and Rueckert, Daniel and Wang, Zehan , title =

    Shi, Wenzhe and Caballero, Jose and Huszar, Ferenc and Totz, Johannes and Aitken, Andrew P. and Bishop, Rob and Rueckert, Daniel and Wang, Zehan , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , month =

  44. [44]

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

    Octnet: Learning deep 3d representations at high resolutions , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  45. [46]

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

    Choy, Christopher and Gwak, JunYoung and Savarese, Silvio , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

  46. [47]

    ACM Trans

    Wang, Peng-Shuai and Liu, Yang and Guo, Yu-Xiao and Sun, Chun-Yu and Tong, Xin , title =. ACM Trans. Graph. , month = jul, articleno =. 2017 , issue_date =. doi:10.1145/3072959.3073608 , abstract =

    work page doi:10.1145/3072959.3073608 2017

  47. [48]

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

    Kpconv: Flexible and deformable convolution for point clouds , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  48. [49]

    Proceedings of the European Conference on Computer Vision (ECCV) , month =

    Jayaraman, Pradeep Kumar and Mei, Jianhan and Cai, Jianfei and Zheng, Jianmin , title =. Proceedings of the European Conference on Computer Vision (ECCV) , month =

  49. [50]

    Proceedings of the 56th Annual IEEE/ACM International Symposium on Microarchitecture , pages =

    Tang, Haotian and Yang, Shang and Liu, Zhijian and Hong, Ke and Yu, Zhongming and Li, Xiuyu and Dai, Guohao and Wang, Yu and Han, Song , title =. Proceedings of the 56th Annual IEEE/ACM International Symposium on Microarchitecture , pages =. 2023 , isbn =. doi:10.1145/3613424.3614303 , abstract =

    work page doi:10.1145/3613424.3614303 2023

  50. [51]

    2023 , eprint=

    Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling , author=. 2023 , eprint=

    2023

  51. [52]

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

    Dai, Jifeng and Qi, Haozhi and Xiong, Yuwen and Li, Yi and Zhang, Guodong and Hu, Han and Wei, Yichen , title =. Proceedings of the IEEE International Conference on Computer Vision (ICCV) , month =

  52. [53]

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

    Lu, Tao and Yu, Mulin and Xu, Linning and Xiangli, Yuanbo and Wang, Limin and Lin, Dahua and Dai, Bo , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2024 , pages =

    2024

  53. [54]

    Submanifold Sparse Convolutional Networks

    Submanifold Sparse Convolutional Networks , author=. arXiv preprint arXiv:1706.01307 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  54. [55]

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

    Sun, Long and Dong, Jiangxin and Tang, Jinhui and Pan, Jinshan , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2023 , pages =

    2023

  55. [56]

    arXiv preprint arXiv:2603.18513 , year=

    CAFlow: Adaptive-Depth Single-Step Flow Matching for Efficient Histopathology Super-Resolution , author=. arXiv preprint arXiv:2603.18513 , year=

    work page arXiv

  56. [57]

    arXiv preprint arXiv:2412.06028 , year=

    Sparsedit: Token sparsification for efficient diffusion transformer , author=. arXiv preprint arXiv:2412.06028 , year=

    work page arXiv

  57. [58]

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

    Peng, Bohao and Wu, Xiaoyang and Jiang, Li and Chen, Yukang and Zhao, Hengshuang and Tian, Zhuotao and Jia, Jiaya , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2024 , pages =

    2024

  58. [59]

    Graph Attention Networks

    Graph attention networks , author=. arXiv preprint arXiv:1710.10903 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  59. [60]

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

    Timofte, Radu and Agustsson, Eirikur and Van Gool, Luc and Yang, Ming-Hsuan and Zhang, Lei , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops , month =

  60. [61]

    Accelerating the Super-Resolution Convolutional Neural Network

    Dong, Chao and Loy, Chen Change and Tang, Xiaoou. Accelerating the Super-Resolution Convolutional Neural Network. Computer Vision -- ECCV 2016. 2016

    2016

  61. [62]

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

    Li, Yawei and Zhang, Kai and Liang, Jingyun and Cao, Jiezhang and Liu, Ce and Gong, Rui and Zhang, Yulun and Tang, Hao and Liu, Yun and Demandolx, Denis and Ranjan, Rakesh and Timofte, Radu and Van Gool, Luc , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops , month =. 2023 , pages =

    2023

  62. [63]

    and Fowlkes, C

    Martin, D. and Fowlkes, C. and Tal, D. and Malik, J. , booktitle=. A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics , year=

  63. [64]

    Multimedia tools and applications , volume=

    Sketch-based manga retrieval using manga109 dataset , author=. Multimedia tools and applications , volume=. 2017 , publisher=

    2017

  64. [65]

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

    Huang, Jia-Bin and Singh, Abhishek and Ahuja, Narendra , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , month =

  65. [66]

    2012 , publisher=

    Low-complexity single-image super-resolution based on nonnegative neighbor embedding , author=. 2012 , publisher=

    2012

  66. [67]

    On Single Image Scale-Up Using Sparse-Representations

    Zeyde, Roman and Elad, Michael and Protter, Matan. On Single Image Scale-Up Using Sparse-Representations. Curves and Surfaces. 2012

    2012

  67. [68]

    and Sheikh, H.R

    Zhou Wang and Bovik, A.C. and Sheikh, H.R. and Simoncelli, E.P. , journal=. Image quality assessment: from error visibility to structural similarity , year=

  68. [69]

    and Shechtman, Eli and Wang, Oliver , title =

    Zhang, Richard and Isola, Phillip and Efros, Alexei A. and Shechtman, Eli and Wang, Oliver , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , month =

  69. [70]

    , journal=

    Ding, Keyan and Ma, Kede and Wang, Shiqi and Simoncelli, Eero P. , journal=. Image Quality Assessment: Unifying Structure and Texture Similarity , year=

  70. [71]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Arbitrary-steps image super-resolution via diffusion inversion , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  71. [72]

    International Conference on Computer Vision Workshops (ICCVW) , date =

    Xintao Wang and Liangbin Xie and Chao Dong and Ying Shan , title =. International Conference on Computer Vision Workshops (ICCVW) , date =

  72. [73]

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

    Jeong, Jinho and Han, Sangmin and Kim, Jinwoo and Kim, Seon Joo , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2025 , pages =

    2025

  73. [74]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    Effective Diffusion Transformer Architecture for Image Super-Resolution , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2025 , month=. doi:10.1609/aaai.v39i3.32247 , abstractNote=

    work page doi:10.1609/aaai.v39i3.32247 2025

  74. [75]

    SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis , url =

    Podell, Dustin and English, Zion and Lacey, Kyle and Blattmann, Andreas and Dockhorn, Tim and M\". SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis , url =. International Conference on Learning Representations , editor =