Pith. sign in

REVIEW 4 major objections 5 minor 61 references

Spatial Lifting for Dense Prediction

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Spatial Lifting claims that copying a 2D image into a third dimension and processing it with a channel-constant 3D U-Net can match or beat conventional 2D dense prediction networks while cutting parameters by over 98%.

desk verdict The z-symmetry argument holds: the replicated input makes output slices identical, so the paper's slice-selection and PQA contributions are vacuous; the empirical results are likely real but the framing is wrong. read the letter →

arxiv 2507.10222 v1 pith:IQGM3ED3 submitted 2025-07-14 cs.CV cs.LGeess.IV

Mingzhi Xu , Yizhe Zhang This is my paper · ORCID
classification cs.CVcs.LGeess.IV
keywords spatialliftingdensepredictionsemanticsegmentationdepthestimation3DU-Netqualityassessmentmodelparameterreductionimplicitregularization
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 proposes Spatial Lifting (SL), a method that takes a standard 2D image, copies it sixteen times along a new third axis, and feeds the resulting 3D volume to a 3D U-Net that keeps only 8 channels at every depth. The paper's central claim is that this counterintuitive dimensionality increase lets the small 3D model match or beat a conventional 2D U-Net with channels growing from 32 to 512, on 13 semantic segmentation datasets and 6 depth estimation datasets. If the claim holds, dense prediction models could shrink by over 98% in parameters and lower inference cost while staying competitive, and the replicated output slices provide a near-free per-image quality score. A sympathetic reading is that SL is offered as a new modeling paradigm: trade channel width for an extra spatial axis and get efficiency plus built-in uncertainty estimation.

What carries the argument

The central object is the lifted volume: an input image replicated $m=16$ times along a new depth axis, producing a 3D tensor that a 3D U-Net processes into $m$ output slice predictions. The key identity is the complexity trade-off: a 2D U-Net's parameters grow like the product of successive channel counts (32×64×...×512), while SL keeps a constant $C_*=8$ channels so parameters scale with $C_*^2$ times the kernel volume, and MACs only pick up a factor $m \cdot K_z$. The third mechanism is slice selection and quality scoring: after training, the $s=5$ slices with the lowest average training loss (Eq. 6-7) are selected and their logits are summed at test time, and the average Dice between selected and unselected slices defines the quality score $Q$.

What would settle it

On a held-out dataset, compute the per-slice Dice of each of the $m=16$ output slices at test time and compare the ranking with the training-loss ranking used to select the five slices; if the selected slices are not among the best test slices, or if averaging all 16 slices beats the selected five, the slice-selection step is not doing the claimed work.

Watch

Extended reading notes

Core claim

On its own terms, the paper discovers that spatial lifting is a viable alternative to channel-widening in dense prediction. A 3D U-Net with constant 8 channels, fed with a 16-slice lifted volume, is reported to match or outperform a 2D U-Net with 32-512 channels on most of the 13 segmentation datasets (e.g., Dice 90.00 vs 87.76 on DSB2018 for the 5L,2Res config) and to improve depth RMSE and δ1 on most of the 6 depth datasets, with over 98% parameter reduction and lower GMACs in the U-Net case. The paper also claims that the lifted output's slice structure enables a prediction-quality score Q—the mean Dice agreement between the five selected slices and the remaining slices—that correlates with true Dice (Pearson r up to 0.92 on FIVES, mostly significant). The theoretical contribution is the argument that 3D convolutions with shared weights along the lifted axis impose an implicit Lipschitz constraint, giving a smoothness bias and improved generalization, while dense supervision across slices acts as a gradient-averaging regularizer.

Load-bearing premise

The load-bearing premise is that the five output slices with the lowest average training loss are also the best slices to average at test time; the paper does not ablate this selection against using all $m$ slices or a fixed subset.

Editorial extensions

If this is right

  • Dense prediction models can get accurate results with tens of thousands of parameters instead of millions, by replacing channel depth with an extra spatial axis.
  • Every SL model outputs a per-image quality estimate at negligible extra cost, which could be used to flag low-confidence predictions in deployment.
  • SL transfers across backbones and decoders: the paper reports gains or parity when applied to HSNet, CASCADE, EMCAD, FastDepth, and MiDaS on most datasets, so it functions as a plug-in design principle, not just a single architecture.
  • The implicit-regularization story predicts that SL models should degrade smoothly under perturbation or distribution shift along the lifted axis, because the shared 3D weights enforce z-smoothness.

Reading between the lines

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

  • The slice-selection step is the least guarded part of the pipeline: if the training-loss ranking of slices does not transfer to test data, an adaptive test-time selection rule (for example, picking the most mutually consistent slices) could be tried instead.
  • The lift factor $m$ is fixed at 16 in all experiments; a scaling study varying $m$ would clarify whether the gains come from having many slices per se or from the channel reduction, and would give practical guidance for new tasks.
  • The same recipe might transfer to other input dimensions, such as lifting video or point clouds into a 4D volume, where the extra axis could act as a pseudo-time or pseudo-scale dimension.
  • Since $Q$ measures cross-slice agreement, it could be turned into a training objective that explicitly enforces slice consistency, which the paper does not currently propose.
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

4 major / 5 minor

Summary. This paper proposes Spatial Lifting (SL) for dense prediction. A 2D image is replicated m times along a new z-axis (Eq. 1) and processed by a 3D U-Net with a constant, small channel width (C*=8). All m output slices are supervised with the same 2D ground truth; at the end of training the s=5 slices with lowest average training loss are selected (Eqs. 6-7), and at test time the selected slice logits are summed (Eq. 9). The authors claim that this scheme outperforms or matches conventional 2D U-Nets with 32-512 channels on 13 semantic segmentation and 6 depth estimation datasets while reducing parameters by over 98% and lowering GMACs, that it produces structured outputs along z that enable a near-free prediction quality assessment (Eq. 11), and that it provides implicit regularization. Theoretical support is offered in Section III-C.

Significance. If the proposed mechanism were real, the paper would be significant: a simple dimensionality-lifting recipe that improves accuracy while cutting parameters by two orders of magnitude, plus a built-in uncertainty signal, would be of broad interest. The empirical evaluation is unusually broad (19 datasets, two tasks, multiple backbones/decoders), and the efficiency tables (Tables II and VI) are clearly presented. However, the central mechanism is called into question by the z-translation equivariance of the architecture, and the paper's own results do not resolve this; the PQA correlations in Table IV cannot be taken as evidence of the proposed mechanism without an explicit symmetry-breaking component or an ablation. The significance therefore hinges on a correction that is not presently in the manuscript.

major comments (4)
  1. [III-A, Eq. (1)] The input I' is constant along the lifted z-axis, and the 3D U-Net used in Section IV-B is composed of 3D convolutions, ReLU, normalization, pooling, and upsampling, none of which break translation symmetry along z away from boundaries. The paper describes no positional encoding, per-slice embedding, or other symmetry-breaking mechanism. Consequently, for the bulk of the m interior slices, the output P(x,y,z,c) is identical, and the only z-variation arises from padding/pooling boundary effects. This makes the slice selection in Eqs. (6)-(7) and the PQA score in Eq. (11) measures of boundary artifacts rather than of the claimed 'structured outputs along the lifted dimension' (Section I, Proposition III.4). The authors must either introduce explicit per-slice position information (e.g., a learned z embedding) or demonstrate empirically that slice predictions differ in a way that reflects the proposed mechanism rather than boundary effects.
  2. [III-A, Eqs. (6)-(7)] The selection of z1..zs as the s slices with lowest average training loss is a post-hoc selection on the training labels. The paper never ablates this choice against random s-slice subsets or against averaging all m slices. If the ranking of slices on training data does not transfer to test data, the numbers in Tables III, VII, and VIII could change substantially; and if, as suggested above, most slices are identical, the selection is vacuous. An ablation comparing training-loss-based selection, random fixed selection, and all-slice averaging is necessary before the test-time protocol can be trusted.
  3. [III-C, Theorems III.1-III.2 and Propositions III.3-III.4] The theoretical analysis does not currently support the empirical claims. Theorem III.1 establishes only a generic Lipschitz property of convolutions, which says nothing specific about the lifted input when that input is constant along z; Theorem III.2 and Proposition III.3 are heuristic analogies to Rademacher complexity and gradient averaging, not formal statements. More importantly, the claimed 'implicit regularization' and 'variance reduction' arguments presuppose that different slices compute different features, which is exactly what the z-translation equivariance issue denies. Proposition III.4, in particular, rests on the assumption of divergent predictions across z, and if most slices are identical the Q statistic cannot behave as described.
  4. [IV-B, Tables II-III] The accuracy comparison is confounded by channel width. The 2D U-Net baselines use channel widths of 32-512, while SL-UNet uses 8 channels throughout. No matched-capacity 2D baseline (e.g., a 2D U-Net with 8 channels and comparable GMACs) is reported, so the observed Dice improvements cannot be attributed to dimensionality lifting per se. Since parameter/compute reduction is the paper's headline, a matched-width or matched-compute ablation is necessary to support the claim that spatial lifting, rather than some other training or architecture difference, is responsible for the results.
minor comments (5)
  1. [III-A, Eq. (9)] For multi-class segmentation, the activation sigma should be a softmax over classes rather than a sigmoid; as written, summing logits and then applying a per-class sigmoid before argmax is not a standard probability combination for single-label segmentation.
  2. [III-B, Eq. (11)] The notation BM(.,.,z,.) is not defined before its use in the PQA score; please define the binarization procedure explicitly.
  3. [IV-A, Table I vs. text] The text states that NYU Depth V2 has 795 training images, while Table I reports 50,688 training images; these numbers need to be reconciled.
  4. [IV-D, Table VIII] The text says the SL models achieve 'significant improvements' across most datasets, but several RMSE entries degrade (e.g., ResNeXt+SL-MiDaS on Make3D and KITTI); please quantify the trade-off and avoid the impression of uniform gains.
  5. [II-D, Fig. 2] The interpretability visualizations in Fig. 2 are not quantitatively connected to slice selection or to the PQA mechanism; consider showing which slices are selected and how their visualizations differ from unselected ones.

Circularity Check

2 steps flagged · score 6.0 of 10

The PQA and 'emergent structure' claims reduce to Eq. (1): with a z-invariant input and a z-translation-equivariant 3D U-Net, interior output slices are identical by construction, making slice selection and prediction-quality assessment vacuous.

  1. self definitional [Section III-A, Eq. (1); Section III-B, Eq. (11); Section III-C.3, Proposition III.4]
    "I′(x, y, z, c) = I(x, y, c), ∀z ∈ {1, 2, . . . , m}, (1) ... The prediction quality score is computed as the average of the Dice similarities across all pairs of selected and unselected slices: Q = 1/(s·(m−s)) Σ ... Dice(BM(·,·,zi,·), BM(·,·,z′,·)). (11)"

    The lifted input is defined to be constant along z (Eq. 1), and the network is an off-the-shelf 3D U-Net whose convolutions, normalizations, nonlinearities, pooling, and upsampling are z-translation equivariant; the paper describes no positional encoding, per-slice weights, or other symmetry-breaking mechanism. Hence every interior output slice P(·,·,z,·) is identical to every other interior slice. Eq. (6)-(7) therefore select among equal training losses, Eq. (9) sums copies of the same logits, and Eq. (11) compares a segmentation map with itself (up to boundary artifacts). The claimed 'emergent structure' and 'prediction quality assessment' are thus not learned; they are present in the input by construction.

  2. renaming known result [Section III-A, Eq. (1) and Section III-D; Section VI Conclusion]
    "SL operates by lifting low-dimensional inputs, such as 2D images, into a higher-dimensional space and processing them with deep networks designed for that dimension, like a 3D U-Net."

    With I′ z-invariant and no symmetry-breaking in the 3D U-Net, a 3D convolution with kernel K(x,y,z) acts exactly as a 2D convolution with kernel K′(x,y) = Σ_z K(x,y,z) on the original image, plus boundary effects at the ends of the replicated axis. 'Spatial Lifting' is therefore a re-description of a standard 2D convolution with an extra summation over a replicated axis, not a genuinely new higher-dimensional modeling paradigm. The claimed novelty is the same computation expressed in lifted coordinates, i.e., a renaming of a known convolutional operation rather than a new mechanism.

full rationale

The benchmark accuracy and efficiency comparisons are externally grounded and are not circular: SL-UNet is evaluated against standard UNet and PVT-based decoders on fixed datasets, and the parameter/MAC reductions follow from the chosen channel counts. The training-set slice selection of Eqs. (6)-(7) is a legitimate, if unablated, model-selection step rather than a fitted parameter relabeled as a prediction. However, the paper's central conceptual claims — 'intrinsically structured outputs along the lifted dimension' and the near-zero-cost prediction quality assessment — reduce to the construction in Eq. (1). Because the input is replicated identically along z and the off-the-shelf 3D U-Net is z-translation equivariant, interior output slices coincide; Eq. (11) therefore measures self-similarity of identical maps, so the reported PQA correlations cannot be produced by the described mechanism. The theoretical Proposition III.4 assumes slice disagreement that the specified architecture cannot generate. This is a partial, construction-level circularity affecting the PQA and 'emergent structure' contributions, while the accuracy/efficiency claims retain independent empirical content. Score 6 rather than 8-10 because the main performance comparisons do not reduce to the circular step.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on several hand-chosen hyperparameters (m=16, s=5, C*=8) and a data-dependent slice selection that uses training labels. The theoretical sections assume standard but unspecified distributional assumptions and rely on proof sketches rather than complete derivations.

free parameters (4)
  • m (lifting depth) = 16
    The number of times the input is replicated along the new axis is fixed at 16; this is chosen by hand and not adapted per dataset (acknowledged in Section V).
  • s (number of selected slices) = 5
    At test time, the 5 slices with lowest average training loss are summed (default s=5, Section III-A). This is a hyperparameter set without a systematic search.
  • C* (channel width of SL networks) = 8
    SL-UNet uses a constant channel width of 8 across all layers, while the baseline UNet uses 32-512 channels. This is a hand-chosen value that largely determines the parameter savings.
  • Slice indices z1..zs = Selected per trained model by lowest average training loss (Eq. 6,7)
    The final prediction uses only a subset of the m slices; which slices are used is determined after training on the training set labels. This is a data-dependent selection rule, not a fixed architecture parameter.
assumptions (4)
  • standard math The (k+1)-D convolutional layer enforces a Lipschitz constraint along the lifted dimension through weight sharing (Theorem III.1).
    Any convolutional layer is Lipschitz; the specific claim about z-dimension smoothness is a direct consequence of weight sharing, but the paper does not prove a quantitative bound.
  • domain assumption Standard assumptions on the data distribution and loss regularity for the generalization bound (Theorem III.2).
    The statement says 'under standard assumptions' but does not specify them, so the improved bound is conditional on unstated conditions.
  • domain assumption A well-trained network produces consistent predictions across slices for in-distribution inputs, and disagreement indicates uncertainty (Proposition III.4).
    This is the heuristic behind the PQA score; it is plausible but not proved, and the empirical correlations are moderate.
  • ad hoc to paper The 3D U-Net with constant low channels can represent the dense prediction task.
    The choice of a narrow 3D U-Net is motivated by efficiency, but there is no analysis of the expressive capacity needed for the target tasks.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Spatial Lifting for Dense Prediction." pith.science (2026). https://pith.science/paper/IQGM3ED3

@misc{pith2026250710222,
  author       = {Pith},
  title        = {Pith review of: Spatial Lifting for Dense Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IQGM3ED3}},
  note         = {Machine review of arXiv:2507.10222}
}
read the original abstract

We present Spatial Lifting (SL), a novel methodology for dense prediction tasks. SL operates by lifting standard inputs, such as 2D images, into a higher-dimensional space and subsequently processing them using networks designed for that higher dimension, such as a 3D U-Net. Counterintuitively, this dimensionality lifting allows us to achieve good performance on benchmark tasks compared to conventional approaches, while reducing inference costs and significantly lowering the number of model parameters. The SL framework produces intrinsically structured outputs along the lifted dimension. This emergent structure facilitates dense supervision during training and enables robust, near-zero-additional-cost prediction quality assessment at test time. We validate our approach across 19 benchmark datasets (13 for semantic segmentation and 6 for depth estimation), demonstrating competitive dense prediction performance while reducing the model parameter count by over 98% (in the U-Net case) and lowering inference costs. Spatial Lifting introduces a new vision modeling paradigm that offers a promising path toward more efficient, accurate, and reliable deep networks for dense prediction tasks in vision.

Figures

Figures reproduced from arXiv: 2507.10222 by the authors.

Figure 1
Figure 1. High-level view of Spatial Lifting for Dense Prediction. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Visualization results using different interpretability methods for each depth slice. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. (a) Semantic segmentation results; (b) Depth estimation results. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 42 canonical work pages

  1. [1]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III

    work page 2015

  2. [2]

    Rethinking atrous convolution for semantic image segmentation,

    L.-C. Chen, G. Papandreou, F. Schroff, and H. Adam, “Rethinking atrous convolution for semantic image segmentation,” arXiv preprint arXiv:1706.05587, 2017

    arXiv 2017

  3. [3]

    Towards robust monocular depth estimation: Mixing datasets for zero-shot cross- dataset transfer,

    R. Ranftl, K. Lasinger, D. Hafner, K. Schindler, and V . Koltun, “Towards robust monocular depth estimation: Mixing datasets for zero-shot cross- dataset transfer,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 3, pp. 1623–1637, 2020

    work page 2020

  4. [4]

    Adabins: Depth estimation using adaptive bins,

    S. F. Bhat, I. Alhashim, and P. Wonka, “Adabins: Depth estimation using adaptive bins,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2021, pp. 4009–4018

    work page 2021

  5. [5]

    Raft: Recurrent all-pairs field transforms for optical flow,

    Z. Teed and J. Deng, “Raft: Recurrent all-pairs field transforms for optical flow,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part II 16 . Springer, 2020, pp. 402–419

    2020

  6. [6]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 770–778

    2016

  7. [7]

    Fully convolutional networks for semantic segmentation,

    J. Long, E. Shelhamer, and T. Darrell, “Fully convolutional networks for semantic segmentation,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2015, pp. 3431–3440

    2015

  8. [8]

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

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

    arXiv 2010

Show all 61 references
  1. [9]

    Segformer: Simple and efficient design for semantic segmentation with transformers,

    E. Xie, W. Wang, Z. Yu, A. Anandkumar, J. M. Alvarez, and P. Luo, “Segformer: Simple and efficient design for semantic segmentation with transformers,” in Advances in Neural Information Processing Systems , vol. 34, 2021, pp. 12 077–12 090

    2021

  2. [10]

    Dropout as a bayesian approximation: Representing model uncertainty in deep learning,

    Y . Gal and Z. Ghahramani, “Dropout as a bayesian approximation: Representing model uncertainty in deep learning,” in International Conference on Machine Learning . PMLR, 2016, pp. 1050–1059

    2016

  3. [11]

    Mobilenets: Efficient convolu- tional neural networks for mobile vision applications,

    A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto, and H. Adam, “Mobilenets: Efficient convolu- tional neural networks for mobile vision applications,” arXiv preprint arXiv:1704.04861, 2017

    2017 arXiv

  4. [12]

    Shufflenet: An extremely efficient convolutional neural network for mobile devices,

    X. Zhang, X. Zhou, M.-H. Lin, and J. Sun, “Shufflenet: An extremely efficient convolutional neural network for mobile devices,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 6848–6856

    2018

  5. [13]

    Efficientnet: Rethinking model scaling for convolutional neural networks,

    M. Tan and Q. V . Le, “Efficientnet: Rethinking model scaling for convolutional neural networks,” in International conference on machine learning. PMLR, 2019, pp. 6105–6114

    2019

  6. [14]

    Learning both weights and connections for efficient neural network,

    S. Han, J. Pool, J. Tran, and W. Dally, “Learning both weights and connections for efficient neural network,” in Advances in neural information processing systems , vol. 28, 2015

    2015

  7. [15]

    Quantizing deep convolutional networks for efficient inference: A whitepaper,

    R. Krishnamoorthi, “Quantizing deep convolutional networks for efficient inference: A whitepaper,” arXiv preprint arXiv:1806.08342 , 2018

    2018 arXiv

  8. [16]

    Training data-efficient image transformers & distillation through attention,

    H. Touvron, M. Cord, M. Douze, F. Massa, A. Sablayrolles, and H. Jégou, “Training data-efficient image transformers & distillation through attention,” in International conference on machine learning . PMLR, 2021, pp. 10 347–10 357

    2021

  9. [17]

    3d u-net: learning dense volumetric segmentation from sparse annotation,

    Ö. Çiçek, A. Abdulkadir, S. S. Lienkamp, T. Brox, and O. Ronneberger, “3d u-net: learning dense volumetric segmentation from sparse annotation,” in Medical Image Computing and Computer-Assisted Intervention– MICCAI 2016: 19th International Conference, Athens, Greece, October 1...

    2016

  10. [18]

    Springer, 2015, pp. 234–241

    2015

  11. [19]

    Pyramid scene parsing network,

    H. Zhao, J. Shi, X. Qi, X. Wang, and J. Jia, “Pyramid scene parsing network,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , July 2017

    2017

  12. [20]

    Spatial pyramid pooling in deep convolutional networks for visual recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Spatial pyramid pooling in deep convolutional networks for visual recognition,” IEEE Transactions on 11 Pattern Analysis and Machine Intelligence, vol. 37, no. 9, pp. 1904–1916, 2015

    1904

  13. [21]

    Rethinking semantic segmentation from a sequence-to-sequence perspective with transformers,

    S. Zheng, J. Lu, H. Zhao, X. Zhu, Z. Luo, Y . Wang, Y . Fu, J. Feng, T. Xi- ang, P. H. Torr, and L. Zhang, “Rethinking semantic segmentation from a sequence-to-sequence perspective with transformers,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Rec...

    2021

  14. [22]

    The lottery ticket hypothesis: Finding sparse, trainable neural networks,

    J. Frankle and M. Carbin, “The lottery ticket hypothesis: Finding sparse, trainable neural networks,” arXiv preprint arXiv:1803.03635 , 2018

    2018 arXiv

  15. [23]

    Distilling the knowledge in a neural network,

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” arXiv preprint arXiv:1503.02531 , 2015

    2015 arXiv

  16. [24]

    Darts: Differentiable architecture search,

    H. Liu, K. Simonyan, and Y . Yang, “Darts: Differentiable architecture search,” arXiv preprint arXiv:1806.09055 , 2018

    2018 arXiv

  17. [25]

    Mnasnet: Platform-aware neural architecture search for mobile,

    M. Tan, B. Chen, R. Pang, V . Vasudevan, M. Sandler, A. Howard, and Q. V . Le, “Mnasnet: Platform-aware neural architecture search for mobile,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 2820–2828

    2019

  18. [26]

    Proxylessnas: Direct neural architecture search on target task and hardware,

    H. Cai, L. Zhu, and S. Han, “Proxylessnas: Direct neural architecture search on target task and hardware,” arXiv preprint arXiv:1812.00332 , 2018

    2018 arXiv

  19. [27]

    Real-time single image and video super- resolution using an efficient sub-pixel convolutional neural network,

    W. Shi, J. Caballero, F. Huszár, J. Totz, A. P. Aitken, R. Bishop, D. Rueckert, and Z. Wang, “Real-time single image and video super- resolution using an efficient sub-pixel convolutional neural network,” 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), ...

    2016

  20. [28]

    de Berg, Computational geometry: algorithms and applications

    M. de Berg, Computational geometry: algorithms and applications . Springer Science & Business Media, 2000

    2000

  21. [29]

    Some fundamental aspects about lipschitz continuity of neural networks,

    G. Khromov and S. P. Singh, “Some fundamental aspects about lipschitz continuity of neural networks,” arXiv preprint arXiv:2302.10886 , 2023

    2023 arXiv

  22. [30]

    Local rademacher complexities,

    P. L. Bartlett, O. Bousquet, and S. Mendelson, “Local rademacher complexities,” Annals of Statistics , vol. 33, pp. 1497–1537, 2005

    2005

  23. [31]

    Pvt v2: Improved baselines with pyramid vision transformer,

    W. Wang, E. Xie, X. Li, D.-P. Fan, K. Song, D. Liang, T. Lu, P. Luo, and L. Shao, “Pvt v2: Improved baselines with pyramid vision transformer,” Computational Visual Media , vol. 8, no. 3, pp. 415–424, 2022

    2022

  24. [32]

    An ensemble classification-based approach applied to retinal blood vessel segmentation,

    M. M. Fraz, P. Remagnino, A. Hoppe, B. Uyyanonvara, A. R. Rudnicka, C. G. Owen, and S. A. Barman, “An ensemble classification-based approach applied to retinal blood vessel segmentation,”IEEE Transactions on Biomedical Engineering , vol. 59, no. 9, pp. 2538–2548, 2012

    2012

  25. [33]

    Nucleus segmentation across imaging experiments: the 2018 data science bowl,

    J. C. Caicedo, A. Goodman, K. W. Karhohs, B. A. Cimini, J. Ackerman, M. Haghighi, C. Heng, T. Becker, M. Doan, C. McQuin, M. H. Rohban, S. Singh, and A. E. Carpenter, “Nucleus segmentation across imaging experiments: the 2018 data science bowl,” Nature Methods, vol. 16, pp. 12...

    2018

  26. [34]

    Kvasir-seg: A segmented polyp dataset,

    D. Jha, P. H. Smedsrud, M. A. Riegler, P. Halvorsen, T. De Lange, D. Johansen, and H. D. Johansen, “Kvasir-seg: A segmented polyp dataset,” in International conference on multimedia modeling . Springer, 2019, pp. 451–462

    2019

  27. [35]

    Monusac2020: A multi-organ nuclei segmentation and classification challenge,

    R. Verma, N. Kumar, A. Patil, N. C. Kurian, S. Rane, S. Graham, Q. D. Vu, M. Zwager, S. E. A. Raza, N. Rajpoot et al., “Monusac2020: A multi-organ nuclei segmentation and classification challenge,” IEEE Transactions on Medical Imaging , vol. 40, no. 12, pp. 3413–3423, 2021

    2021

  28. [36]

    N. C. Codella, D. Gutman, M. E. Celebi, B. Helba, M. A. Marchetti, S. W. Dusza, A. Kalloo, K. Liopyris, N. Mishra, H. Kittler et al., “Skin lesion analysis toward melanoma detection: A challenge at the 2017 international symposium on biomedical imaging (isbi), hosted by the in...

    2017

  29. [37]

    The ham10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions,

    P. Tschandl, C. Rosendahl, and H. Kittler, “The ham10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions,” Scientific Data, vol. 5, no. 1, pp. 1–9, 2018

    2018

  30. [38]

    A dataset and a technique for generalized nuclear segmentation for computational pathology,

    N. Kumar, R. Verma, S. Sharma, S. Bhargava, A. Vahadane, and A. Sethi, “A dataset and a technique for generalized nuclear segmentation for computational pathology,” IEEE Transactions on Medical Imaging , vol. 36, no. 7, pp. 1550–1560, 2017

    2017

  31. [39]

    Sartorius - cell instance segmentation,

    A. Howard, A. Chow, CorporateResearchSartorius, M. Ca, P. Culliton, and T. Jackson, “Sartorius - cell instance segmentation,” https://kaggle. com/competitions/sartorius-cell-instance-segmentation, 2021, accessed: 2025-05-24

    2021

  32. [40]

    Segmentation of nuclei in histopathology images by deep regression of the distance map,

    P. Naylor, M. Laé, F. Reyal, and T. Walter, “Segmentation of nuclei in histopathology images by deep regression of the distance map,” IEEE Transactions on Medical Imaging , vol. 38, no. 2, pp. 448–459, 2018

    2018

  33. [41]

    Neural control of fasting-induced torpor in mice,

    T. Hitrec, M. Luppi, S. Bastianini, F. Squarcio, C. Berteotti, V . Lo Martire, D. Martelli, A. Occhinegro, D. Tupone, G. Zoccoli et al., “Neural control of fasting-induced torpor in mice,” Scientific Reports, vol. 9, no. 1, p. 15462, 2019

    2019

  34. [42]

    Automating cell counting in fluorescent microscopy through deep learning with c-resunet,

    R. Morelli, L. Clissa, R. Amici, M. Cerri, T. Hitrec, M. Luppi, L. Rinaldi, F. Squarcio, and A. Zoccoli, “Automating cell counting in fluorescent microscopy through deep learning with c-resunet,” Scientific Reports , vol. 11, no. 1, p. 22920, 2021

    2021

  35. [43]

    Gland segmentation in colon histology images: The glas challenge contest,

    K. Sirinukunwattana, J. P. Pluim, H. Chen, X. Qi, P.-A. Heng, Y . B. Guo, L. Y . Wang, B. J. Matuszewski, E. Bruni, U. Sanchez et al. , “Gland segmentation in colon histology images: The glas challenge contest,” Medical Image Analysis , vol. 35, pp. 489–502, 2017

    2017

  36. [44]

    A stochastic polygons model for glandular structures in colon histology images,

    K. Sirinukunwattana, D. R. Snead, and N. M. Rajpoot, “A stochastic polygons model for glandular structures in colon histology images,” IEEE Transactions on Medical Imaging , vol. 34, no. 11, pp. 2366–2378, 2015

    2015

  37. [45]

    Blood cell segmentation dataset,

    D. S. Depto, S. Rahman, M. M. Hosen, M. S. Akter, T. R. Reme, A. Rahman, H. Zunai, M.R.C.Mahdy, M. S. Rahman, and J. B. Lahiri, “Blood cell segmentation dataset,” https://www.kaggle.com/dsv/6107556, 2023, accessed: 2025-05-24

    2023

  38. [46]

    Fives: A fundus image dataset for artificial intelligence based vessel segmentation,

    K. Jin, X. Huang, J. Zhou, Y . Li, Y . Yan, Y . Sun, Q. Zhang, Y . Wang, and J. Ye, “Fives: A fundus image dataset for artificial intelligence based vessel segmentation,” Scientific Data, vol. 9, no. 1, p. 475, 2022

    2022

  39. [47]

    Nuinsseg: a fully annotated dataset for nuclei instance segmentation in h&e-stained histological images,

    A. Mahbod, C. Polak, K. Feldmann, R. Khan, K. Gelles, G. Dorffner, R. Woitek, S. Hatamikia, and I. Ellinger, “Nuinsseg: a fully annotated dataset for nuclei instance segmentation in h&e-stained histological images,” Scientific Data, vol. 11, no. 1, p. 295, 2024

    2024

  40. [48]

    The cityscapes dataset for semantic urban scene understanding,

    M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benenson, U. Franke, S. Roth, and B. Schiele, “The cityscapes dataset for semantic urban scene understanding,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 3213–3223

    2016

  41. [49]

    End-to-end multi-task learning with attention,

    S. Liu, E. Johns, and A. J. Davison, “End-to-end multi-task learning with attention,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2019, pp. 1871–1880

    2019

  42. [50]

    Learning depth from single monocular images,

    A. Saxena, S. Chung, and A. Ng, “Learning depth from single monocular images,” Advances in neural information processing systems , vol. 18, 2005

    2005

  43. [51]

    Learning 3-d scene structure from a single still image,

    A. Saxena, M. Sun, and A. Y . Ng, “Learning 3-d scene structure from a single still image,” in 2007 IEEE 11th international conference on computer vision. IEEE, 2007, pp. 1–8

    2007

  44. [52]

    Diode: A dense indoor and outdoor depth dataset,

    I. Vasiljevic, N. I. Kolkin, S. Zhang, R. Luo, H. Wang, F. Z. Dai, A. F. Daniele, M. Mostajabi, S. Basart, M. R. Walter, and G. Shakhnarovich, “Diode: A dense indoor and outdoor depth dataset,” ArXiv, vol. abs/1908.00463, 2019

    1908 arXiv

  45. [53]

    Vision meets robotics: The kitti dataset,

    A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, “Vision meets robotics: The kitti dataset,” The international journal of robotics research , vol. 32, no. 11, pp. 1231–1237, 2013

    2013

  46. [54]

    Indoor segmentation and support inference from rgbd images,

    N. Silberman, D. Hoiem, P. Kohli, and R. Fergus, “Indoor segmentation and support inference from rgbd images,” in Computer Vision–ECCV 2012: 12th European Conference on Computer Vision, Florence, Italy, October 7-13, 2012, Proceedings, Part V 12 . Springer, 2012, pp. 746– 760

    2012

  47. [55]

    Modest museum dataset,

    A. Shan, “Modest museum dataset,” https://www.kaggle.com/datasets/ shanwizard/modest-museum-dataset, 2021, accessed: 2025-05-25

    2021

  48. [56]

    Hsnet: A hybrid semantic network for polyp segmentation,

    W. Zhang, C. Fu, Y . Zheng, F. Zhang, Y . Zhao, and C.-W. Sham, “Hsnet: A hybrid semantic network for polyp segmentation,” Computers in Biology and Medicine , vol. 150, p. 106173, 2022

    2022

  49. [57]

    Emcad: Efficient multi- scale convolutional attention decoding for medical image segmentation,

    M. M. Rahman, M. Munir, and R. Marculescu, “Emcad: Efficient multi- scale convolutional attention decoding for medical image segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2024, pp. 11 769–11 779

    2024

  50. [58]

    Medical image segmentation via cascaded attention decoding,

    M. M. Rahman and R. Marculescu, “Medical image segmentation via cascaded attention decoding,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , January 2023, pp. 6222–6231

    2023

  51. [59]

    Mobilenetv2: Inverted residuals and linear bottlenecks,

    M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “Mobilenetv2: Inverted residuals and linear bottlenecks,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 4510–4520

    2018

  52. [60]

    Aggregated residual transformations for deep neural networks,

    S. Xie, R. Girshick, P. Dollár, Z. Tu, and K. He, “Aggregated residual transformations for deep neural networks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2017, pp. 1492–1500

    2017

  53. [61]

    Fastdepth: Fast monocular depth estimation on embedded systems,

    D. Wofk, F. Ma, T.-J. Yang, S. Karaman, and V . Sze, “Fastdepth: Fast monocular depth estimation on embedded systems,” in 2019 International Conference on Robotics and Automation (ICRA) . IEEE, 2019, pp. 6101– 6108

    2019

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.