Pith. sign in

REVIEW 3 major objections 5 minor 49 references

HBONet: Harmonious Bottleneck on Two Orthogonal Dimensions

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper claims that replacing MobileNetV2's bottlenecks with a nested spatial-and-channel HBO unit yields higher accuracy at equal or lower FLOPs, with gains growing as the budget shrinks, reaching 6.6% top-1 on ImageNet.

desk verdict A solid, incremental architecture paper whose reported gains are real but not yet causally tied to the named HBO mechanism; send it to review but demand an ablation. read the letter →

arxiv 1908.03888 v1 pith:7DWJSALP submitted 2019-08-11 cs.CV cs.LGeess.IV

Duo Li , Aojun Zhou , Anbang Yao This is my paper · ORCID
classification cs.CVcs.LGeess.IV
keywords HarmoniousBottlenecklightweightCNNdepthwiseseparableconvolutionspatialfeaturescaleMobileNetV2efficientarchitectureImageclassificationobjectdetection
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 establish that a modest architecture unit can make extremely lightweight convolutional networks substantially more accurate without raising their cost. The unit, the Harmonious Bottleneck on two Orthogonal dimensions (HBO), wraps the usual channel expansion-and-contraction inside a spatial contraction-and-expansion, so the expensive wide middle features are computed at reduced resolution. On the paper's evidence, swapping MobileNetV2's bottlenecks for HBO blocks lifts ImageNet top-1 accuracy by up to 6.6% at under 40 MFLOPs, with similar jumps on PASCAL VOC detection and Market-1501 person re-identification. If true, this pushes the practical accuracy floor for low-power vision systems, where small accuracy losses traditionally translate into large performance drops.

What carries the argument

The load-bearing object is the HBO block: an inverted channel bottleneck ($1\times1$ expansion, $3\times3$ depthwise convolution, $1\times1$ linear contraction) nested inside spatial contraction and expansion. The contraction is a strided $5\times5$ depthwise convolution that maps $h \times w \times c_1$ to $h/s \times w/s \times c_1$; the expansion is bilinear upsampling followed by a $5\times5$ depthwise convolution. A residual connection and a concatenation of half the output channels from the input or its pooled version are included. The arithmetic that pays for the extra structure is direct: if the inserted channel block costs $B$, wrapping it in the spatial pair makes the total cost $B/s^2 + (h/s \times w/s \times c_1 + h \times w \times c_2)\,k^2$, so the wide middle features are computed at a fraction of the usual price.

What would settle it

Run two ImageNet experiments matched in FLOPs, width, expansion factor, projection, and shortcut design—one with HBO blocks and one with identical blocks whose main branch stays at full spatial resolution instead of contracting and expanding. If the full-resolution version matches HBO's accuracy, the spatial contraction-expansion component is not the source of the reported gains.

Watch

Extended reading notes

Core claim

On the paper's own terms, the contribution is a bottleneck that treats spatial feature size and feature depth as two coordinated axes rather than independent design choices. A standard MobileNetV2 inverted residual expands channels, applies a depthwise convolution at full spatial scale, and contracts back. HBO instead first shrinks the spatial map with a strided depthwise convolution, runs the channel expansion and contraction on the small map, then upsamples back to the original size; a residual path and a concatenation of half the channels from the input (or a pooled version) round out the block. Because channel width and spatial size vary in opposite directions inside the block, the expensive wide feature maps live on a cheap small grid. The paper reports that HBONets built this way beat re-implemented MobileNetV2 baselines at matched complexity across five width multipliers, with the margin growing as the budget falls: 0.9 points at 300 MFLOPs, 5.0 points at 37 MFLOPs, and 6.6 points at 14 MFLOPs for ImageNet.

Load-bearing premise

The central claim credits HBO's nested spatial and channel transformations for the gains, but the compared networks also differ in channel widths, expansion factor, projection layers, and concatenation paths, so the specific contribution of the HBO structure itself is not isolated.

Editorial extensions

If this is right

  • The accuracy advantage over MobileNetV2 grows as the width multiplier falls, so the module is most useful precisely where lightweight deployment is hardest: under roughly 40 MFLOPs.
  • The gains transfer from classification to detector backbones: on PASCAL VOC with SSD, HBONet (0.1) improves mAP by 6.3 points over MobileNetV2 (0.1) at the same 320x320 input.
  • For person re-identification on Market-1501, the narrowest HBONet raises Rank-1 accuracy by 5.0 points, indicating the learned representation is not overfit to class labels.
  • At the larger end, HBONet (1.0) reports 73.1% ImageNet top-1 at 305 MFLOPs, above the listed efficient architectures around 300 MFLOPs.
  • Stacking several spatial contraction units inside one block and increasing the upsampling rate can push accuracy further, as the 4x variant in the paper's Table 6 shows.

Reading between the lines

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

  • The paper never isolates the spatial contraction-expansion from accompanying changes in channel widths, expansion factor, projection layers, and concatenation paths, so a fair reader can attribute part of the gain to those modifications; a factorial ablation is the natural next test.
  • The input-concatenation trick resembles feature reuse in dense connectivity and may contribute independently of the spatial nesting; varying only that path would quantify its effect.
  • Because the spatial wrapper only assumes a strided downsampling followed by an upsampling, the same wrapper could be applied to other channel-centric blocks such as ShuffleNet-style group convolutions, though the paper does not test this.
  • The reported trend of larger gains at smaller budgets suggests the module's benefit may be a complexity-preserving regularizer rather than purely added capacity; one testable extension is whether HBO blocks still win at higher budgets when all other modifications are held fixed.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes the Harmonious Bottleneck (HBO), a building block for lightweight CNNs that couples a spatial contraction-expansion operation with a channel expansion-contraction operation in a nested, bilaterally symmetric structure. The authors replace MobileNetV2 bottlenecks with HBO blocks to construct HBONets and evaluate them on ImageNet classification, PASCAL VOC object detection, and Market-1501 person re-identification. They report consistent accuracy gains over MobileNetV2 at matched or lower FLOPs, with the largest margins (up to 6.6% top-1 on ImageNet, 6.3 mAP on VOC, and 5.0 Rank-1 on Market-1501) at very low computational budgets below 40 MFLOPs. The paper also presents a cascade variant with multiple successive spatial contraction units and reports a modest additional improvement.

Significance. If the causal claim is substantiated, the HBO module would be a useful addition to the lightweight-architecture toolbox, particularly for the sub-40 MFLOPs regime where the reported gains are substantial. The empirical contribution is strengthened by the breadth of evaluation across three very different tasks and five width multipliers, by the matched-FLOPs comparisons, and by the public release of code and pretrained models. The most likely source of practical value is the architecture itself: HBONet (1.0) reaches 73.1% top-1 at ~300 MFLOPs, competitive with or better than several contemporary efficient networks. However, the paper's central interpretational claim — that the spatial contraction-expansion mechanism, rather than the accompanying width, expansion-factor, projection, and concatenation changes, drives the gains — is not yet demonstrated by the experiments as designed.

major comments (3)
  1. [§3.4, Tables 2–4] The central claim that the reported gains come from the HBO spatial contraction-expansion structure is not supported by the experiments as presented. Section 3.4 states that "some other modifications are also made instead of performing a trivial replacement": per-layer widths are re-tuned; the expansion factor t is lowered from 6 to 2 for HBO blocks; a 1x1 projection convolution is inserted between block groups; and half of the output channels are concatenated from the input or its pooled version. Every HBONet in Tables 2–4 differs from its MobileNetV2 baseline in all of these dimensions simultaneously, so the accuracy margins (up to 6.6%, 6.3%, and 5.0%) cannot be attributed to the HBO module alone. The authors should provide a controlled comparison that replaces MobileNetV2 bottlenecks with HBO blocks while holding width, expansion factor, projection, and concatenation fixed, and then ablate each of these modifications separately.
  2. [Table 6, Fig. 4] The cascade variant experiment does not include a k=1 (or s=1) no-spatial-contraction control. Comparing HBONet(2x), HBONet(4x), and HBONet(8x) shows a non-monotone pattern (58.3, 59.3, 58.2), but without the k=1 baseline it cannot be determined whether any of these numbers reflects the spatial contraction mechanism rather than the accompanying capacity rebalancing and concatenation changes. Moreover, the text notes that the HBONet(0.25) channels here are set divisible by 8, unlike the Table 2 configuration, so the comparison is not anchored to the main result. A k=1 control at identical channel settings is necessary to support the claim that spatial contraction units provide further improvement.
  3. [§4.1, Tables 2–9] All accuracy numbers are reported without training variance. Many of the headline margins, especially the 0.9-point gain at width 1.0 and the smaller detection/re-ID gaps, are within the run-to-run variability typically observed for ImageNet training with a fixed seed set; reporting a single run makes the comparisons, and especially the monotonic trend in Table 2, difficult to assess. At minimum, the authors should report results over multiple seeds or provide evidence that the training procedure is stable (e.g., by showing that the reproduced MobileNetV2 numbers are close to official baselines across runs).
minor comments (5)
  1. [§4.1] In the training description, "Stochastic Gradient Descend" should read "Stochastic Gradient Descent."
  2. [References] Reference [42] contains a stray comma in the author list ("Jianghuang Lai,"), and the author list appears incomplete; please verify against the original publication.
  3. [Figures 2 and 3] The two figures use different kernel sizes for the depthwise convolutions in the spatial contraction path (5x5 in Figure 2 versus 3x3 in Figure 3). If the 5x5 is intentional, please state so explicitly and explain how it interacts with the stride and upsampling; otherwise, unify the figures.
  4. [Table 6 caption] The caption says a variant is denoted HBONet (2k×) if the maximum number of spatial contraction units is k, but the table rows use HBONet (2×), (4×), and (8×). Please make the notation consistent and clarify whether k counts the number of stride-2 units or the total downsampling factor.
  5. [§3.2] The phrase "bilaterally symmetric structure" is evocative but not formally defined. Please specify the precise symmetry condition between the spatial contraction-expansion and channel expansion-contraction components, or rephrase to avoid an undefined structural claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: HBONet's reported gains are external benchmark measurements, not constructed from its own definitions or fitted inputs.

full rationale

The paper contains no circular derivation in the sense defined here. The central claim is that replacing MobileNetV2 bottlenecks with HBO modules yields accuracy gains at comparable or lower FLOPs. These gains are measured on ImageNet, PASCAL VOC, and Market-1501 under fixed training protocols, so the outcome is not defined by an input parameter, nor is any fitted value renamed as a prediction. The HBO module is defined structurally (spatial contraction-expansion and channel expansion-contraction), and the reported accuracies are independent empirical results. The paper's self-citations (e.g., references [47] and [48] on quantization work by overlapping authors) appear only in the related-work discussion and are not load-bearing for the architecture claim. There is no invoked uniqueness theorem, no ansatz justified solely by a self-citation, and no equation in which the predicted accuracy is equivalent by construction to a fitted quantity. The absence of an ablation that fully isolates the spatial contraction component from the accompanying width, expansion-factor, projection, and concatenation changes is a legitimate concern about internal validity and causal attribution, but it is not circularity: the performance numbers would stand even if the attribution were wrong. Accordingly, the appropriate finding is no significant circularity.

Assumptions & free parameters 4 free parameters · 4 assumptions · 1 invented entities

The central claim is empirical, so the ledger mainly lists hand-chosen architectural hyperparameters and standard domain assumptions. No speculative entities are introduced beyond the module itself.

free parameters (4)
  • Expansion factor t for HBO blocks = 2
    Set lower than MobileNetV2's value of 6 to keep FLOPs comparable; no ablation over t is reported.
  • Spatial downsampling rate s = 2 (default), 4 and 8 in variants
    Controls compute savings; variants in Table 6 show small differences but the default is used in the main comparison.
  • Half-channel concatenation ratio = 0.5 (half of output channels from input)
    Reduces the channels the main branch must compute; not ablated.
  • Width multiplier configurations = 1.0, 0.8, 0.5, 0.35, 0.25, 0.1
    Chosen to match MobileNetV2 FLOPs; the gain magnitude depends on these choices.
assumptions (4)
  • domain assumption Depthwise separable convolutions provide a valid approximation of standard convolutions for building efficient networks.
    Basis of MobileNetV2 and the HBO module; no theoretical proof given.
  • domain assumption Bilinear upsampling and strided downsampling inside a block do not harm training or add significant computational cost.
    Used in the spatial contraction-expansion component; not separately verified.
  • domain assumption The MobileNetV2 baselines reproduced by the authors are strong and representative; the official low-width baselines are not suboptimal.
    The reported gains, especially at width 0.1, depend on the baseline quality.
  • domain assumption Single-crop accuracy on ImageNet and mAP/Rank-1 on the other benchmarks are sufficient to measure the accuracy-efficiency trade-off.
    Standard practice in this literature.
invented entities (1)
  • Harmonious Bottleneck (HBO) module independent evidence
    purpose: A bottleneck unit combining spatial contraction-expansion with channel expansion-contraction to improve accuracy at low FLOPs.
    The module is the paper's contribution; its effect is empirically testable and the authors release code for replication.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HBONet: Harmonious Bottleneck on Two Orthogonal Dimensions." pith.science (2026). https://pith.science/paper/7DWJSALP

@misc{pith2026190803888,
  author       = {Pith},
  title        = {Pith review of: HBONet: Harmonious Bottleneck on Two Orthogonal Dimensions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7DWJSALP}},
  note         = {Machine review of arXiv:1908.03888}
}
read the original abstract

MobileNets, a class of top-performing convolutional neural network architectures in terms of accuracy and efficiency trade-off, are increasingly used in many resourceaware vision applications. In this paper, we present Harmonious Bottleneck on two Orthogonal dimensions (HBO), a novel architecture unit, specially tailored to boost the accuracy of extremely lightweight MobileNets at the level of less than 40 MFLOPs. Unlike existing bottleneck designs that mainly focus on exploring the interdependencies among the channels of either groupwise or depthwise convolutional features, our HBO improves bottleneck representation while maintaining similar complexity via jointly encoding the feature interdependencies across both spatial and channel dimensions. It has two reciprocal components, namely spatial contraction-expansion and channel expansion-contraction, nested in a bilaterally symmetric structure. The combination of two interdependent transformations performing on orthogonal dimensions of feature maps enhances the representation and generalization ability of our proposed module, guaranteeing compelling performance with limited computational resource and power. By replacing the original bottlenecks in MobileNetV2 backbone with HBO modules, we construct HBONets which are evaluated on ImageNet classification, PASCAL VOC object detection and Market-1501 person re-identification. Extensive experiments show that with the severe constraint of computational budget our models outperform MobileNetV2 counterparts by remarkable margins of at most 6.6%, 6.3% and 5.0% on the above benchmarks respectively. Code and pretrained models are available at https://github.com/d-li14/HBONet.

Figures

Figures reproduced from arXiv: 1908.03888 by the authors.

Figure 1
Figure 1. Comparison of top-1 accuracy and FLOPs for Mo [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Bottleneck comparison between MobileNetV2 and HBONet. The lower design represents our proposed HBO module while the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Basic modules of Inverted Residual with Linear Bottleneck and Harmonious Bottleneck, shown as (1) and (2) respectively. Each [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Schema of variant blocks including k successive spatial [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 28 canonical work pages

  1. [1]

    Fast patch-based style transfer of arbitrary style

    Tian Qi Chen and Mark Schmidt. Fast patch-based style transfer of arbitrary style. arXiv preprint arXiv:1612.04337, 2016

    arXiv 2016

  2. [2]

    Xception: Deep learning with depthwise separable convolutions

    Francois Chollet. Xception: Deep learning with depthwise separable convolutions. In CVPR, 2017

    2017

  3. [3]

    Mark Everingham, S. M. Ali Eslami, Luc Van Gool, Christo- pher K. I. Williams, John Winn, and Andrew Zisserman. The pascal visual object classes challenge: A retrospective.Inter- national Journal of Computer Vision, 111(1):98–136, 2015

    work page 2015

  4. [4]

    Rich feature hierarchies for accurate object detection and semantic segmentation

    Ross Girshick, Jeff Donahue, Trevor Darrell, and Jitendra Malik. Rich feature hierarchies for accurate object detection and semantic segmentation. In CVPR, 2014

    2014

  5. [5]

    Deep learning with limited numerical precision

    Suyog Gupta, Ankur Agrawal, Kailash Gopalakrishnan, and Pritish Narayanan. Deep learning with limited numerical precision. In ICML, 2015

    2015

  6. [6]

    Learn- ing both weights and connections for efficient neural net- works

    Song Han, Jeff Pool, John Tran, and William J Dally. Learn- ing both weights and connections for efficient neural net- works. In NIPS, 2015

    work page 2015

  7. [7]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR, 2016

    2016

  8. [8]

    Channel pruning for accelerating very deep neural networks

    Yihui He, Xiangyu Zhang, and Jian Sun. Channel pruning for accelerating very deep neural networks. In ICCV, 2017

    work page 2017

Show all 49 references
  1. [9]

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

    Andrew G Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco An- dreetto, and Hartwig Adam. Mobilenets: Efficient convolu- tional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861, 2017

    2017 arXiv

  2. [10]

    Squeeze-and-excitation net- works

    Jie Hu, Li Shen, and Gang Sun. Squeeze-and-excitation net- works. In CVPR, 2018

    2018

  3. [11]

    Condensenet: An efficient densenet using learned group convolutions

    Gao Huang, Shichen Liu, Laurens van der Maaten, and Kil- ian Q Weinberger. Condensenet: An efficient densenet using learned group convolutions. In CVPR, 2018

    2018

  4. [12]

    Densely connected convolutional net- works

    Gao Huang, Zhuang Liu, Laurens van der Maaten, and Kil- ian Q Weinberger. Densely connected convolutional net- works. In CVPR, 2017

    2017

  5. [13]

    Speed/accuracy trade-offs for modern convolutional object detectors

    Jonathan Huang, Vivek Rathod, Chen Sun, Menglong Zhu, Anoop Korattikara, Alireza Fathi, Ian Fischer, Zbigniew Wo- jna, Yang Song, Sergio Guadarrama, and Kevin Murphy. Speed/accuracy trade-offs for modern convolutional object detectors. In CVPR, 2017

    2017

  6. [14]

    Binarized neural networks

    Itay Hubara, Matthieu Courbariaux, Daniel Soudry, Ran El- Yaniv, and Yoshua Bengio. Binarized neural networks. In NIPS, 2016

    2016

  7. [15]

    Squeezenet: Alexnet-level accuracy with 50x fewer parame- ters and <0.5mb model size

    Forrest N Iandola, Song Han, Matthew W Moskewicz, Khalid Ashraf, William J Dally, and Kurt Keutzer. Squeezenet: Alexnet-level accuracy with 50x fewer parame- ters and <0.5mb model size. In ICLR, 2017

    2017

  8. [16]

    Batch normalization: Accelerating deep network training by reducing internal co- variate shift

    Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep network training by reducing internal co- variate shift. In ICML, 2015

    2015

  9. [17]

    Flattened convolutional neural networks for feedforward ac- celeration

    Jonghoon Jin, Aysegul Dundar, and Eugenio Culurciello. Flattened convolutional neural networks for feedforward ac- celeration. arXiv preprint arXiv:1412.5474, 2015

    2015 arXiv

  10. [18]

    Imagenet classification with deep convolutional neural net- works

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural net- works. In NIPS, 2012

    2012

  11. [19]

    Pruning filters for efficient convnets

    Hao Li, Asim Kadav, Igor Durdanovic, Hanan Samet, and Hans Peter Graf. Pruning filters for efficient convnets. In ICLR, 2017

    2017

  12. [20]

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C. Berg. Ssd: Single shot multibox detector. In ECCV, 2016

    2016

  13. [21]

    Fully convolutional networks for semantic segmentation

    Jonathan Long, Evan Shelhamer, and Trevor Darrell. Fully convolutional networks for semantic segmentation. In CVPR, 2015

    2015

  14. [22]

    Shufflenet v2: Practical guidelines for efficient cnn architec- ture design

    Ningning Ma, Xiangyu Zhang, Hai-Tao Zheng, and Jian Sun. Shufflenet v2: Practical guidelines for efficient cnn architec- ture design. In ECCV, 2018

    2018

  15. [23]

    Stacked hour- glass networks for human pose estimation

    Alejandro Newell, Kaiyu Yang, and Jia Deng. Stacked hour- glass networks for human pose estimation. In ECCV, 2016

    2016

  16. [24]

    Automatic differentiation in pytorch

    Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan, Edward Yang, Zachary DeVito, Zeming Lin, Al- ban Desmaison, Luca Antiga, and Adam Lerer. Automatic differentiation in pytorch. In NIPS-W, 2017

    2017

  17. [25]

    Accelerating deep neural net- works with spatial bottleneck modules

    Junran Peng, Lingxi Xie, Zhaoxiang Zhang, Tieniu Tan, and Jingdong Wang. Accelerating deep neural net- works with spatial bottleneck modules. arXiv preprint arXiv:1809.02601, 2018

    2018 arXiv

  18. [26]

    Efficient neural architecture search via parameter sharing

    Hieu Pham, Melody Y Guan, Barret Zoph, Quoc V Le, and Jeff Dean. Efficient neural architecture search via parameter sharing. In ICML, 2018

    2018

  19. [27]

    Z. Qin, Z. Zhang, X. Chen, C. Wang, and Y . Peng. Fd- mobilenet: Improved mobilenet with a fast downsampling strategy. In ICIP, 2018

    2018

  20. [28]

    Unsuper- vised representation learning with deep convolutional gener- ative adversarial networks

    Alec Radford, Luke Metz, and Soumith Chintala. Unsuper- vised representation learning with deep convolutional gener- ative adversarial networks. In ICLR, 2016

    2016

  21. [29]

    Xnor-net: Imagenet classification using bi- nary convolutional neural networks

    Mohammad Rastegari, Vicente Ordonez, Joseph Redmon, and Ali Farhadi. Xnor-net: Imagenet classification using bi- nary convolutional neural networks. In ECCV, 2016

    2016

  22. [30]

    Reddi, Satyen Kale, and Sanjiv Kumar

    Sashank J. Reddi, Satyen Kale, and Sanjiv Kumar. On the convergence of adam and beyond. In ICLR, 2018

    2018

  23. [31]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In MICCAI, 2015

    2015

  24. [32]

    Imagenet large scale visual recognition challenge

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, San- jeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C Berg, and Fei-Fei Li. Imagenet large scale visual recognition challenge. International Journal of Computer Vision , 1...

    2015

  25. [33]

    Mobilenetv2: Inverted residuals and linear bottlenecks

    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zh- moginov, and Liang-Chieh Chen. Mobilenetv2: Inverted residuals and linear bottlenecks. In CVPR, 2018

    2018

  26. [34]

    Rigid-motion scattering for image classifica- tion

    Laurent Sifre. Rigid-motion scattering for image classifica- tion. Ph.D. thesis, 2014

    2014

  27. [35]

    Very deep convo- lutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convo- lutional networks for large-scale image recognition. InICLR, 2015

    2015

  28. [36]

    Igcv3: Interleaved low-rank group convolutions for efficient deep neural networks

    Ke Sun, Mingjie Li, Dong Liu, and Jingdong Wang. Igcv3: Interleaved low-rank group convolutions for efficient deep neural networks. In BMVC, 2018

    2018

  29. [37]

    Going deeper with convolutions

    Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. Going deeper with convolutions. In CVPR, 2015

    2015

  30. [38]

    Rethinking the inception architecture for computer vision

    Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, and Jon Shlens. Rethinking the inception architecture for computer vision. In CVPR, 2016

    2016

  31. [39]

    Im- proving the speed of neural networks on cpus

    Vincent Vanhoucke, Andrew Senior, and Mark Z Mao. Im- proving the speed of neural networks on cpus. In Deep Learning and Unsupervised Feature Learning Workshop, NIPS, 2011

    2011

  32. [40]

    Factorized convolutional neural networks

    Min Wang, Baoyuan Liu, and Hassan Foroosh. Factorized convolutional neural networks. In ICCV, 2017

    2017

  33. [41]

    Wang, Xiang Li, and Charles X

    Robert J. Wang, Xiang Li, and Charles X. Ling. Pelee: A real-time object detection system on mobile devices. In NIPS, 2018

    2018

  34. [42]

    Interleaved structured sparse convolutional neural networks

    Guotai Xie, Jingdong Wang, Ting Zhang, Jianghuang Lai, , Ruichang Hong, and Guo-Jun Qi. Interleaved structured sparse convolutional neural networks. In CVPR, 2018

    2018

  35. [43]

    Aggregated residual transformations for deep neural networks

    Saining Xie, Ross Girshick, Piotr Doll ´ar, Zhuowen Tu, and Kaiming He. Aggregated residual transformations for deep neural networks. In CVPR, 2017

    2017

  36. [44]

    In- terleaved group convolutions

    Ting Zhang, Guo-Jun Qi, Bin Xiao, and Jingdong Wang. In- terleaved group convolutions. In ICCV, 2017

    2017

  37. [45]

    Shufflenet: An extremely efficient convolutional neural net- work for mobile devices

    Xiangyu Zhang, Xinyu Zhou, Mengxiao Lin, and Jian Sun. Shufflenet: An extremely efficient convolutional neural net- work for mobile devices. In CVPR, 2018

    2018

  38. [46]

    Scalable person re-identification: A benchmark

    Liang Zheng, Liyue Shen, Lu Tian, Shengjin Wang, Jing- dong Wang, and Qi Tian. Scalable person re-identification: A benchmark. In ICCV, 2015

    2015

  39. [47]

    Incremental network quantization: Towards lossless cnns with low-precision weights

    Aojun Zhou, Anbang Yao, Yiwen Guo, Lin Xu, and Yurong Chen. Incremental network quantization: Towards lossless cnns with low-precision weights. In ICLR, 2017

    2017

  40. [48]

    Explicit loss-error-aware quantization for low-bit deep neu- ral networks

    Aojun Zhou, Anbang Yao, Kuan Wang, and Yurong Chen. Explicit loss-error-aware quantization for low-bit deep neu- ral networks. In CVPR, 2018

    2018

  41. [49]

    Neural architecture search with reinforcement learning

    Barret Zoph and Quoc V Le. Neural architecture search with reinforcement learning. In ICLR, 2017

    2017

Pith tools

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