ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
Open accessThe DOI matches its Crossref record; 60 of the 60 references it cites were found in a registry or an index; 21 of the 35 sources its claims cite were checked for support.
Abstract
This work proposes to evaluate the direct metric on the target platform, beyond only considering FLOPs, and derives several practical guidelines for efficient network design, called ShuffleNet V2.
Overview
CoChat Summary
TL;DR
The paper argues that CNN architecture design should optimize direct speed/latency on the target platform rather than only FLOPs, because memory access cost, parallelism, and platform-specific optimization also determine runtime. It derives four practical guidelines and introduces ShuffleNet V2, which achieves better speed–accuracy tradeoffs than ShuffleNet v1, MobileNet v2, Xception, and others on GPU and ARM.
Key findings
- Four guidelines are proposed: equal channel width minimizes memory access cost (MAC); excessive group convolution increases MAC; network fragmentation reduces parallelism; element-wise operations are non-negligible. Validation shows, for example, that using 8 groups is more than 2× slower than 1 group on GPU and up to 30% slower on ARM, and a 4-fragment structure is 3× slower than a 1-fragment structure on GPU.
- ShuffleNet V2 outperforms prior lightweight models: at 40M FLOPs it is 3.5% and 3.7% more accurate than ShuffleNet v1 and MobileNet v2, respectively. At 500M FLOPs it is 58% faster than MobileNet v2, 63% faster than ShuffleNet v1, and 25% faster than Xception on GPU.
- On ImageNet classification, ShuffleNet V2 models report top-1 error / GPU speed / ARM speed of 39.7% / 417 / 57.0 at 41 MFLOPs (0.5×), 30.6% / 341 / 24.4 at 146 MFLOPs (1×), 27.4% / 255 / 11.8 at 299 MFLOPs (1.5×), and 25.1% / 217 / 6.7 at 591 MFLOPs (2×).
- On COCO object detection, ShuffleNet V2 achieves mmAP of 22.5, 29.0, 31.8, and 33.3 at 40M, 140M, 300M, and 500M FLOPs, respectively, and the ShuffleNet V2* variant reaches 23.7, 29.6, 32.2, and 34.2. For large models, ShuffleNet v2-50 at 2.3G FLOPs has 22.8% top-1 error versus 25.2% for ShuffleNet v1-50 and 24.0% for ResNet-50 at 3.8G FLOPs.
Methods
- Controlled runtime experiments on two platforms: a single NVIDIA GeForce GTX 1080Ti with CUDNN 7.0, and a Qualcomm Snapdragon 810 with a Neon-based single-thread implementation.
- Ablation benchmarks isolate design factors: channel ratios, group numbers, fragmented vs. non-fragmented blocks, and ReLU/shortcut element-wise operations, with FLOPs fixed across variants where applicable.
- Architecture evaluation uses ImageNet 2012 classification at roughly 40, 140, 300, and 500+ MFLOPs, plus COCO object detection with Light-Head RCNN, comparing ShuffleNet V2 against ShuffleNet v1, MobileNet v2, Xception, DenseNet, CondenseNet, IGCV2/IGCV3, and auto-generated models.
Limitations
- Stated: the MAC analysis assumes cache is large enough to store entire feature maps and parameters, so real MAC may deviate from the theoretical bound; the guidelines are platform-independent but validated only on GPU and ARM, and the paper argues the direct metric should be evaluated on the target platform.
- Stated: COCO detection is not tested on ARM because PSRoI Pooling is unavailable there; speed measurements for CondenseNets are omitted due to lack of an efficient implementation; the reimplemented MobileNet v2 results are not as good as those reported in the original paper.
- Evident from the text: evaluation is limited to ImageNet classification and COCO detection on two hardware platforms, with specific batch sizes and single-thread ARM testing, so generalization to other devices, tasks, and deployment settings is not established.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
4 Valid · 7 Not valid · 8 Not checked · 58 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Recent advances in light-weight neural network architectures [15,13,14,9,11,10,12] are mostly based on the metric of FLOPs and do not consider these properties above.”
What the sources say
- Not matchedcites [15] 1. Krizhevsky, A., Sutskever, I., Hinton, G.E.: Imagenet classification with deep convolutional neural networks. In: Advances in neural information processing systems. (2012) 1097-1105
Its entry has no title and year that we can search by.
- Could not verifycites [13] 31. Ioffe, S., Szegedy, C.: Batch normalization: Accelerating deep network training by reducing internal covariate shift. In: International conference on machine learning. (2015) 448-456 sourceExperimental — accuracy not yet measured
The Conclusion (p. 7) presents batch normalization as a way to accelerate training, not as a survey of lightweight architectures. The Inception-model appendix (p. 9) discusses computational cost for one architecture variant, but does not say that recent lightweight architectures are mostly based on FLOPs or that they neglect the unspecified properties. Searches for FLOPs and lightweight architectures found no passage addressing that claim. (LLM, conf=0.75)
- Not checked — no full text of the sourcecites [14] 20. Jaderberg, M., Vedaldi, A., Zisserman, A.: Speeding up convolutional neural networks with low rank expansions. arXiv preprint arXiv:1405.3866 (2014) source
- Partly supportedcites [9] 8. Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. arXiv preprint arXiv:1709.01507 (2017) sourceExperimental — accuracy not yet measured
Related Work (p. 2) says much prior research focused on reducing model and computational complexity, and §4 (p. 4) compares architectures using GFLOPs. But §4 also examines runtime and parameter counts; the paper does not establish that lightweight architectures are mostly based on FLOPs or that they neglect the unspecified ‘properties above.’ (LLM, conf=0.76)
- Partly supportedcites [11] 6. Huang, G., Liu, Z., Weinberger, K.Q., van der Maaten, L.: Densely connected convolutional networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. Volume 1. (2017) 3 sourceExperimental — accuracy not yet measured
The Conclusion (p. 8) describes DenseNet as requiring fewer parameters and less computation, and Figure 3 (p. 6) compares performance by FLOPs. But the paper does not say lightweight architectures are mostly based on FLOPs; its Introduction (p. 2) also emphasizes information flow and gradient propagation. The unspecified “properties above” cannot be identified from the citing sentence. (LLM, conf=0.83)
- Partly supportedcites [10] 16. Huang, G., Liu, S., van der Maaten, L., Weinberger, K.Q.: Condensenet: An efficient densenet using learned group convolutions. arXiv preprint arXiv:1711.09224 (2017) sourceExperimental — accuracy not yet measured
The Introduction describes CondenseNet as an efficient architecture, and the ImageNet comparison evaluates it using FLOPs. But the paper also reports parameter counts and actual inference time on an ARM processor (Comparison with state-of-the-art efficient CNNs; Actual inference time). It therefore supports CondenseNet as an example of a lightweight architecture evaluated by FLOPs, not the broader assertion that such architectures disregard other properties. No page numbers are available in our copy. (LLM, conf=0.84)
- Partly supportedcites [12] 25. Ioannou, Y., Robertson, D., Cipolla, R., Criminisi, A.: Deep roots: Improving cnn efficiency with hierarchical filter groups. arXiv preprint arXiv:1605.06489 (2016) sourceExperimental — accuracy not yet measured
The Introduction (p. 1) identifies this as work on reducing CNN computational cost and model size, so it supports citing the paper as an example of a lightweight architecture. But the Conclusion (p. 8) evaluates efficiency using CPU and GPU timings as well as parameter counts, and GPU Implementation (p. 7) notes that runtime gains differed from FLOPs-based predictions. The copy therefore does not support characterizing this work as mostly based on FLOPs; the unspecified “properties above” cannot be assessed from the citing sentence. (LLM, conf=0.83)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Second, operations with the same FLOPs could have different running time, depending on the platform.”
What the sources say
- Partly supportedcites [19] 17. Liu, Z., Li, J., Shen, Z., Huang, G., Yan, S., Zhang, C.: Learning efficient convolutional networks through network slimming. In: 2017 IEEE International Conference on Computer Vision (ICCV), IEEE (2017) 2755-2763 sourceExperimental — accuracy not yet measured
The Introduction (page unavailable) indicates that hardware affects inference speed, noting that a large CNN may take minutes to process an image on a mobile device and that some techniques need special hardware for speedup. The Additional information section reports FLOP and wall-clock savings on one GPU, but does not compare equal-FLOP operations across platforms; that part of the claim is unsupported. (LLM, conf=0.76)
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Due to the overheads of data copying (the resolution is as high as 800 × 1200) and other detection-specific operations (like PSRoI Pooling [34]), the speed gap between different models is smaller than that of classification.”
What the sources say
- Could not verifycites [34] 26. Zhang, T., Qi, G.J., Xiao, B., Wang, J.: Interleaved group convolutions for deep neural networks. In: International Conference on Computer Vision. (2017) sourceExperimental — accuracy not yet measured
Our copy of Interleaved Group Convolutions is entirely a classification paper: Sections 5.2-5.4 (pp. 6-8) report CIFAR-10, CIFAR-100, SVHN and ImageNet classification accuracy/parameter/FLOP comparisons, and no passage addresses object detection, PSRoI/RoI pooling, data copying at 800x1200 resolution, or a comparison of speed gaps between detection and classification models. Searches run on 'detection / PSRoI / RoI pooling / Faster R-CNN / mAP', 'speed gap classification', 'data copying 800x1200 resolution running time' and 'inference time GPU' returned only classification tables and references. The sole practical-speed remark in our copy is in 5.4 (p. 8) — that NVIDIA CuDNN did not yet support group convolutions, making group convolution slow in implementation — which neither supports nor contradicts the claim. No work is named in our copy as the origin of the detection-speed point, so misattribution cannot be established either. (LLM, conf=0.72)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Second, in each block, half of feature channels (when c ′ = c/ 2) directly go through the block and join the next block. This can be regarded as a kind of feature reuse , in a similar spirit as in DenseNet [6] and CondenseNet [16].”
What the sources say
- Could not verifycites [6] 5. He, K., Zhang, X., Ren, S., Sun, J.: Identity mappings in deep residual networks. In: European Conference on Computer Vision, Springer (2016) 630-645 sourceExperimental — accuracy not yet measured
The Introduction (p. 1) and Analysis of Deep Residual Networks (p. 2) describe identity shortcuts that let information pass directly between ResNet units. Searches for ShuffleNet V2, half-channel bypasses, and feature reuse found no passage addressing the claimed ShuffleNet V2 block design. (LLM, conf=0.76)
- Could not verifycites [16] 34. Li, Z., Peng, C., Yu, G., Zhang, X., Deng, Y., Sun, J.: Light-head r-cnn: In defense of two-stage object detector. arXiv preprint arXiv:1711.07264 (2017) sourceExperimental — accuracy not yet measured
The Introduction (p. 2) and §3.1.2 (p. 3) describe thin feature maps for a two-stage detector, not ShuffleNet V2 blocks. Searches for half-channel bypasses and feature reuse found no passage addressing the claimed block design; the References (p. 9) list an earlier ShuffleNet paper but do not describe this feature. (LLM, conf=0.77)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
None of the sources cited for this claim have been checked yet.
In the paper
“As shown in Figure 2, in light-weight models like [15,14], element-wise operations occupy considerable amount of time, especially on GPU.”
What the sources say
- Not matchedcites [15] 1. Krizhevsky, A., Sutskever, I., Hinton, G.E.: Imagenet classification with deep convolutional neural networks. In: Advances in neural information processing systems. (2012) 1097-1105
Its entry has no title and year that we can search by.
- Not checked — no full text of the sourcecites [14] 20. Jaderberg, M., Vedaldi, A., Zisserman, A.: Speeding up convolutional neural networks with low rank expansions. arXiv preprint arXiv:1405.3866 (2014) source
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
28 more claims
Create a free account to see all 77 claims and to verify the 1 reference still open.
- It may state two separate claims as one. 36
- Probably not a substantial claim. 11
- The quote only partly states this claim. 1
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“they proposed ShuffleNetV2 [120] based on these four guidelines and improved the defects of ShuffleNetV1 by introducing channel splitting.”
From A survey of model compression strategies for object detection (2023) · cites this paper as [120]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 4 Experiment · p. 12
“while on detection the rank becomes ShuffleNet v2 > Xception ≥ ShuffleNet v1 ≥ MobileNet v2.”
§ 4 Experiment · p. 13
“| ShuffleNet v1 | 20.9 | 27.0 | 29.9 | 32.9 | 152 | 85 | 76 | 60 | | MobileNet v2 | 20.7 | 24.4 | 30.0 | 30.6 | 146 | 111 | 94 | 72 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 3 ShuffleNet V2: an Efficient Architecture · p. 9
“In ShuffleNet V2, it is easy to prove that the number of 'directly-connected' channels between i -th and ( i + j )-th building block is r j c , where r = (1 -c ′ ) /c .”
§ 3 ShuffleNet V2: an Efficient Architecture · p. 9
“For simplicity, we set c ′ = c/ 2.”
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 5×
“For validation, we experimented with the “bottleneck” unit (1 × 1 conv followed by 3×3 conv followed by 1×1 conv, with ReLU and shortcut connection) in ResNet [5].”
BackgroundMethodInfluentialSelf-citationVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 1×
“Since the milestone work of AlexNet [15], the ImageNet classification accuracy has been significantly improved by novel structures, including VGG [25], GoogLeNet [28], ResNet [5,6], DenseNet [11], ResNeXt [33], SE-Net [9], and automatic neutral architecture search [18,21,39], to name a few.”
MethodImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 1×
“Since the milestone work of AlexNet [15], the ImageNet classification accuracy has been significantly improved by novel structures, including VGG [25], GoogLeNet [28], ResNet [5,6], DenseNet [11], ResNeXt [33], SE-Net [9], and automatic neutral architecture search [18,21,39], to name a few.”
MethodImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333 · cited 1×
“Our ablation experiments are performed on ImageNet 2012 classification dataset [4,23].”
MethodDensely Connected Convolutional Networks
Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.243 · Cited by 48,566 · cited 9×
“In DenseNet [11], to analyze the feature reuse pattern, the l1-norm of the weights between layers are plotted, as in Fig.”
BackgroundMethodResultInfluentialGoing deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562 · cited 2×
“Since the milestone work of AlexNet [15], the ImageNet classification accuracy has been significantly improved by novel structures, including VGG [25], GoogLeNet [28], ResNet [5,6], DenseNet [11], ResNeXt [33], SE-Net [9], and automatic neutral architecture search [18,21,39], to name a few.”
MethodMicrosoft COCO: Common Objects in Context
Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Dollár, P., & Zitnick, C. L. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10602-1_48 · Cited by 45,460 · cited 1×
“To evaluate the generalization ability, we also tested COCO object detection [17] task.”
MethodImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 1×
“Our ablation experiments are performed on ImageNet 2012 classification dataset [4,23].”
MethodSqueeze-and-Excitation Networks
Hu, J., Shen, L., & Sun, G. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00745 · Cited by 35,667 · cited 5×
“When equipped with Squeezeand-excitation (SE) module [9], the classification accuracy of ShuffleNet v2 is improved by 0.”
BackgroundMethodInfluentialRethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.308 · Cited by 32,138 · cited 1×
“In the GoogLeNet series [27, 29, 28, 13] and auto-generated architectures [39, 21, 18]), a “multi-path” structure is widely adopted in each network block.”
BackgroundMobileNetV2: Inverted Residuals and Linear Bottlenecks
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L.-C. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00474 · Cited by 28,755
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S., & Szegedy, C. (2015) · arXiv · doi:10.48550/arxiv.1502.03167 · Cited by 27,818 · cited 1×
“In the GoogLeNet series [27, 29, 28, 13] and auto-generated architectures [39, 21, 18]), a “multi-path” structure is widely adopted in each network block.”
BackgroundXception: Deep Learning with Depthwise Separable Convolutions
Chollet, F. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.195 · Cited by 20,380 · cited 8×
“Recent advances in light-weight neural network architectures [2,8,18,21,24, 35,39] are mostly based on the metric of FLOPs and do not consider these properties above.”
BackgroundMethodInfluentialMobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017) · arXiv · doi:10.48550/arxiv.1704.04861 · Cited by 19,015 · cited 7×
“Recent advances in light-weight neural network architectures [2,8,18,21,24, 35,39] are mostly based on the metric of FLOPs and do not consider these properties above.”
BackgroundMethodResultInfluentialAggregated Residual Transformations for Deep Neural Networks
Xie, S., Girshick, R., Dollar, P., Tu, Z., & He, K. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.634 · Cited by 12,465 · cited 2×
“Since the milestone work of AlexNet [15], the ImageNet classification accuracy has been significantly improved by novel structures, including VGG [25], GoogLeNet [28], ResNet [5,6], DenseNet [11], ResNeXt [33], SE-Net [9], and automatic neutral architecture search [18,21,39], to name a few.”
BackgroundMethodSqueeze-and-Excitation Networks
Hu, J., Shen, L., Albanie, S., Sun, G., & Wu, E. (2020) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2019.2913372 · Cited by 12,157
Identity mappings in deep residual networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_38 · Cited by 10,741 · cited 5×
“For validation, we experimented with the “bottleneck” unit (1 × 1 conv followed by 3×3 conv followed by 1×1 conv, with ReLU and shortcut connection) in ResNet [5].”
MethodInfluentialSelf-citationIdentity Mappings in Deep Residual Networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · arXiv · doi:10.48550/arxiv.1603.05027 · Cited by 10,665
Self-citationShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices
Zhang, X., Zhou, X., Lin, M., & Sun, J. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00716 · Cited by 9,982 · cited 10×
“1 In this paper, the definition of FLOPs follows [35], i.”
BackgroundMethodInfluentialSelf-citationShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices
Zhang, X., Zhou, X., Lin, M., & Sun, J. (2017) · arXiv · doi:10.48550/arxiv.1707.01083 · Cited by 9,982
Self-citationLearning Transferable Architectures for Scalable Image Recognition
Zoph, B., Vasudevan, V., Shlens, J., & Le, Q. V. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00907 · Cited by 7,831 · cited 8×
“Recent advances in light-weight neural network architectures [2,8,18,21,24, 35,39] are mostly based on the metric of FLOPs and do not consider these properties above.”
BackgroundMethodInfluentialLearning Transferable Architectures for Scalable Image Recognition
Zoph, B., Vasudevan, V., Shlens, J., & Le, Q. V. (2017) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.48550/arxiv.1707.07012 · Cited by 6,390
Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning
Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. (2017) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v31i1.11231 · Cited by 4,636 · cited 1×
“In the GoogLeNet series [27, 29, 28, 13] and auto-generated architectures [39, 21, 18]), a “multi-path” structure is widely adopted in each network block.”
BackgroundDeep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2015) · arXiv · doi:10.48550/arxiv.1512.03385 · Cited by 4,196
Self-citationNeural Architecture Search with Reinforcement Learning
Zoph, B., & Le, Q. V. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1611.01578 · Cited by 3,974 · cited 1×
“Recently, automatic model search [18,21,22,32,38,39] has become a promising trend for CNN architecture design.”
MethodRegularized Evolution for Image Classifier Architecture Search
Real, E., Aggarwal, A., Huang, Y.-P., & Le, Q. V. (2018) · AAAI Conference on Artificial Intelligence · doi:10.48550/arxiv.1802.01548 · Cited by 3,545
Learning Efficient Convolutional Networks through Network Slimming
Liu, Z., Li, J., Shen, Z., Huang, G., Yan, S., & Zhang, C. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.298 · Cited by 2,761 · cited 1×
“Such discrepancy has been noticed in previous works [7,19,24,30].”
ResultImageNet: A large-scale hierarchical image database
Deng, J. H., Dong, W. F., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvprw.2009.5206848 · Cited by 2,714
Channel Pruning for Accelerating Very Deep Neural Networks
He, Y., Zhang, X., & Sun, J. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.155 · Cited by 2,647 · cited 2×
“However, the recent work [7] finds that the decomposition in [36] is even slower on GPU although it reduces FLOPs by 75%.”
BackgroundResultSelf-citationMicrosoft COCO: Common Objects in Context
Lin, T.-Y., Maire, M., Belongie, S., Bourdev, L., Girshick, R., Hays, J., Perona, P., Ramanan, D., Zitnick, C. L., & Dollár, P. (2014) · arXiv · doi:10.48550/arxiv.1405.0312 · Cited by 2,579
MobileNetV2: Inverted Residuals and Linear BottlenecksPreprint
Sandler, M., Howard, A., Zhu, M., et al. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1801.04381 · Cited by 2,255
Squeeze-and-Excitation Networks
Hu, J., Shen, L., Albanie, S., Sun, G. Z., & Wu, E. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1709.01507 · Cited by 1,908
Large Kernel Matters — Improve Semantic Segmentation by Global Convolutional Network
Peng, C., Zhang, X., Yu, G., Luo, G., & Sun, J. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.189 · Cited by 1,750 · cited 1×
“This motivates a practical question: how to increase the size of receptive field? This is critical for object detection in high-resolution images [20].”
BackgroundSelf-citationGoing Deeper with Convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2014) · arXiv · doi:10.48550/arxiv.1409.4842 · Cited by 1,461
Speeding up Convolutional Neural Networks with Low Rank Expansions
Jaderberg, M., Vedaldi, A., & Zisserman, A. (2014) · Proceedings of the British Machine Vision Conference 2014 · doi:10.5244/c.28.88 · Cited by 1,164 · cited 1×
“For example, tensor decomposition is widely used in early works [14,36,37] to accelerate the matrix multiplication.”
cuDNN: Efficient Primitives for Deep Learning
Chetlur, S., Woolley, C., Vandermersch, P., Cohen, J., Tran, J. J., Catanzaro, B., & Shelhamer, E. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1410.0759 · Cited by 943
CondenseNet: An Efficient DenseNet Using Learned Group Convolutions
Huang, G., Liu, S., Maaten, L. V. D., & Weinberger, K. Q. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00291 · Cited by 904 · cited 5×
“The recent CondenseNet [10] also supports the viewpoint.”
BackgroundResultInfluentialAccelerating Very Deep Convolutional Networks for Classification and Detection
Zhang, X., Zou, J., He, K., & Sun, J. (2016) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2015.2502579 · Cited by 832 · cited 2×
“However, the recent work [7] finds that the decomposition in [36] is even slower on GPU although it reduces FLOPs by 75%.”
BackgroundSelf-citationRethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2015) · arXiv · doi:10.48550/arxiv.1512.00567 · Cited by 741
Xie, L., & Yuille, A. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.154 · Cited by 740 · cited 1×
“Recently, automatic model search [18,21,22,32,38,39] has become a promising trend for CNN architecture design.”
MethodImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2014) · arXiv · doi:10.48550/arxiv.1409.0575 · Cited by 567
Xception: Deep Learning with Depthwise Separable Convolutions
Chollet, F. (2016) · arXiv · doi:10.48550/arxiv.1610.02357 · Cited by 538
Large-Scale Evolution of Image Classifiers
Real, E., Moore, S., Selle, A. P., Saxena, S., Suematsu, Y. I. L., Tan, J., Le, Q. V., & Kurakin, A. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1703.01041 · Cited by 466
Speeding up Convolutional Neural Networks with Low Rank Expansions
Jaderberg, M., Vedaldi, A., & Zisserman, A. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1405.3866 · Cited by 447
Learning Structured Sparsity in Deep Neural Networks
Wen, W. X., Wu, C., Wang, Y., Chen, Y., & Li, H. H. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1608.03665 · Cited by 402 · cited 1×
“Such discrepancy has been noticed in previous works [7,19,24,30].”
ResultRegularized Evolution for Image Classifier Architecture Search
Real, E., Aggarwal, A., Huang, Y., & Le, Q. V. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.33014780 · Cited by 328 · cited 5×
“Recent advances in light-weight neural network architectures [2,8,18,21,24, 35,39] are mostly based on the metric of FLOPs and do not consider these properties above.”
BackgroundMethodInfluentialInterleaved Group Convolutions
Zhang, T., Qi, G.-J., Xiao, B., & Wang, J. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.469 · Cited by 323
Deep Roots: Improving CNN Efficiency with Hierarchical Filter Groups
Ioannou, Y., Robertson, D., Cipolla, R., & Criminisi, A. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.633 · Cited by 321 · cited 1×
“Group convolution is at the core of modern network architectures [33, 35, 12, 34, 31, 26].”
BackgroundLight-Head R-CNN: In Defense of Two-Stage Object Detector
Li, Z., Peng, C., Yu, G., Zhang, X., Deng, Y., & Sun, J. (2017) · arXiv · doi:10.48550/arxiv.1711.07264 · Cited by 292
Self-citationEfficient and accurate approximations of nonlinear convolutional networks
Zhang, X., Zou, J., Ming, X., He, K., & Sun, J. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298809 · Cited by 262 · cited 1×
“For example, tensor decomposition is widely used in early works [14,36,37] to accelerate the matrix multiplication.”
Self-citationProgressive Neural Architecture Search
Liu, C., Zoph, B., Neumann, M., Shlens, J., Hua, W., Li, L.-J., Fei-Fei, L., Yuille, A., Huang, J., & Murphy, K. (2017) · arXiv · doi:10.48550/arxiv.1712.00559 · Cited by 171
Distributed Deep Learning Using Synchronous Stochastic Gradient Descent
Das, D., Avancha, S., Mudigere, D., Vaidynathan, K., Sridharan, S., Kalamkar, D. D., Kaul, B., & Dubey, P. K. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1602.06709 · Cited by 138
IGCV3: Interleaved Low-Rank Group Convolutions for Efficient Deep Neural NetworksPreprint
Sun, K., Li, M., Liu, D., & Wang, J. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1806.00178 · Cited by 133
Interleaved Structured Sparse Convolutional Neural Networks
Xie, G., Wang, J., Zhang, T., Lai, J., Hong, R., & Qi, G.-J. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00922 · Cited by 130
CondenseNet: An Efficient DenseNet using Learned Group ConvolutionsPreprint
Huang, G. M., Liu, S., van der Maaten, L., et al. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1711.09224 · Cited by 95
Genetic CNNPreprint
Xie, L., & Yuille, A. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1703.01513 · Cited by 76
Interleaved Group Convolutions for Deep Neural Networks
Zhang, T., Qi, G.-J., Xiao, B., & Wang, J. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1707.02725 · Cited by 73 · cited 1×
“Group convolution is at the core of modern network architectures [33, 35, 12, 34, 31, 26].”
BackgroundLarge Kernel Matters -- Improve Semantic Segmentation by Global Convolutional NetworkPreprint
Peng, C., Zhang, X., Yu, G., et al. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1703.02719 · Cited by 28
Self-citationIGCV$2$: Interleaved Structured Sparse Convolutional Neural NetworksPreprint
Xie, G., Wang, J., Zhang, T., Lai, J., Hong, R., & Qi, G.-J. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1804.06202 · Cited by 24
Deep Roots: Improving CNN Efficiency with Hierarchical Filter Groups
Ioannou, Y., Robertson, D., Cipolla, R., et al. (2016) · arXiv · doi:10.48550/arxiv.1605.06489
Cited by
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
Tan, M., & Le, Q. V. (2019) · arXiv · doi:10.48550/arxiv.1905.11946 · Cited by 17,072
YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors
Wang, C.-Y., Bochkovskiy, A., & Liao, H.-Y. M. (2022) · arXiv · doi:10.48550/arxiv.2207.02696 · Cited by 13,205
YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors
Wang, C.-Y., Bochkovskiy, A., & Liao, H.-Y. M. (2023) · 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52729.2023.00721 · Cited by 13,198
ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks
Wang, Q., Wu, B., Zhu, P., Li, P., Zuo, W., & Hu, Q. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.01155 · Cited by 9,641
“The two most widely used efficient convolutions are group convo-lutions (Zhang et al. 2017; Xie et al. 2017; Ioannou et al. 2017) and depth-wise separable convolutions (Chollet 2017; Sandler et al. 2018; Zhang et al. 2018; Ma et al. 2018).”
BackgroundMethodInfluentialCoordinate Attention for Efficient Mobile Network Design
Hou, Q., Zhou, D., & Feng, J. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.01350 · Cited by 6,961
“ShuffleNetV2 [27] uses a channel split module and a channel shuffle module before and after the inverted residual block.”
Method
Metadata
- Authors
- Ma, N., Zhang, X., Zheng, H.-T., & Sun, J.
- Affiliations
- Megvii (China) (CN)
- Tsinghua University (CN)
- Corresponding author
- Ma, Ningning
- Zhang, Xiangyu
- Venue
- Lecture notes in computer science
- Publisher
- Springer International Publishing
- Published
- 2018
- Pages
- 122-138
- ISSN
- 0302-9743 (ISSN-L)
- 1611-3349
- OpenAlex
- W2883780447
- Cited by
- 7,356
- References
- 39
- Citations by year
- 2026: 804 · 2025: 1317 · 2024: 1402 · 2023: 1328 · 2022: 972 · 2021: 753 · 2020: 518 · 2019: 251 · 2018: 11
- FWCI
- 115.52
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- www.springer.com/tdm (tdm)
- www.springer.com/tdm (vor)
- Locations
- Lecture notes in computer science (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Natural Science Foundation of China (61773229)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
- Keywords
- ShuffleNetV2, speed-accuracy trade-off, inference speed, memory access cost, lightweight neural networks
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.