AS-MLP: An Axial Shifted MLP Architecture for Vision

Lian, Dongze; Yu, Zehao; Sun, Xing; Gao, Shenghua

doi:10.48550/arxiv.2107.08391

Cited by 42 publications

(75 citation statements)

References 46 publications

(55 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Recently, various variants are developed to achieve a better trade-off between accuracy and computational cost. For example, shift operation is introduced in S 2 -MLP [37] and AS-MLP [18] to exchange information across different tokens. Hire-MLP [8] present a hierarchical rearrangement operation, where the inner-region rearrangement and cross-region rearrangement capture local information and global context, respectively.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

PyramidTNT: Improved Transformer-in-Transformer Baselines with Pyramid Architecture

Han¹,

Guo²,

Wang³

2022

Preprint

View full text Add to dashboard Cite

Transformer networks have achieved great progress for computer vision tasks. Transformer-in-Transformer (TNT) architecture utilizes inner transformer and outer transformer to extract both local and global representations. In this work, we present new TNT baselines by introducing two advanced designs: 1) pyramid architecture, and 2) convolutional stem. The new "PyramidTNT" significantly improves the original TNT by establishing hierarchical representations. PyramidTNT achieves better performances than the previous state-of-the-art vision transformers such as Swin Transformer. We hope this new baseline will be helpful to the further research and application of vision transformer. Code will be available at https: //github.com/huawei-noah/CV-Backbones/ tree/master/tnt_pytorch.

show abstract

Section: Related Workmentioning

confidence: 99%

“…In addition, we adopt another setting following [22,18,8], i.e., multi-scale training strategy and "3x" schedule, based on Mask R-CNN [12] and Cascade Mask R-CNN [1].…”

Section: Object Detectionmentioning

confidence: 99%

PyramidTNT: Improved Transformer-in-Transformer Baselines with Pyramid Architecture

Han¹,

Guo²,

Wang³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…This newly proposed network is called S2MLPv2 [94]. Unlike grouping and then performing the same shift operation for each group, the Axial Shifted MLP (AS-MLP) [95] performs different operations within each group which contains a few layers, like 3 or 5. For example, every three consecutive feature maps in the channel direction are, respectively, left-shifted, no-shifted, right-shifted, and so on.…”

Section: Yu Et Al From Baidumentioning

confidence: 99%

“…Specially, the whole architecture contains four stages, where the feature resolution reduces from H/4 × W/4 to H/32 × W/32 and the output dimension increases accordingly. The network based on this design includes Sparse MLP [91], HireMLP [100], AS-MLP [95] and CycleMLP [98]. Patch embedding can be equivalently achieved by a convolution layer with kernel size equal to stride equal to patch size.…”

Section: From Single-stage To Pyramidmentioning

confidence: 99%

See 1 more Smart Citation

Are we ready for a new paradigm shift? A Survey on Visual Deep MLP

Liu¹,

Li²,

Tao³

et al. 2021

Preprint

View full text Add to dashboard Cite

Multilayer perceptron (MLP), as the first neural network structure to appear, was a big hit. But constrained by the hardware computing power and the size of the datasets, it once sank for tens of years. During this period, we have witnessed a paradigm shift from manual feature extraction to the CNN with local receptive field, and further to the Transformer with global receptive field based on selfattention mechanism. And this year (2021), with the introduction of MLP-Mixer, MLP has re-entered the limelight and has attracted extensive research from the computer vision community. Compare to the conventional MLP, it gets deeper but changes the input from full flattening to patch flattening. Given its high performance and less need for vision-specific inductive bias, the community can't help but wonder, Will deep MLP, the simplest structure with global receptive field but no attention, become a new computer vision paradigm? To answer this question, this survey aims to provide a comprehensive overview of the recent development of deep MLP models in vision. Specifically, we review these MLPs in detail, from the subtle sub-module design to the global network structure. We compare the receptive field, computational complexity, and other properties of different network designs in order to understand the development path of MLPs clearly. The investigation shows that MLPs' resolution-sensitivity and computational densities remain unresolved, and pure MLPs are gradually evolving towards CNN-like. We suggest that the current data volume and computational power are not ready to embrace pure MLPs, and artificial visual guidance remains important. Finally, we provide our viewpoint about open research directions and potential future works. We hope this effort will ignite further interest in the community and encourage better visual tailored design for the neural network in the future.

show abstract

SRP&PASMLP‐Net: Lightweight skin lesion segmentation network based on structural re‐parameterization and parallel axial shift multilayer perceptron

Wei,

Chen,

Zhao

et al. 2023

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Accurate skin lesion segmentation (SLS) plays an essential role in the computer‐aided diagnosis of skin diseases, for example, melanomas. However, the automated SLS is challenging due to variations in the nature of skin diseases, particularly the ambiguities of boundaries of skin lesion areas (SLAs) and occlusions by hairs. Though many deep learning models, represented by UNet, have been successfully applied to the SLS over the past years, most suffer from inaccurate SLA segmentation with heavy model parameters and slow inferencing speed, hindering their practical applications. To address these challenges, a lightweight SLS network based on structural re‐parameterization and parallel axial shift multilayer perceptron (MLP), called SRP&PASMLP‐Net, is proposed to achieve powerful segmentation performance with fast inference ability. Specifically, a kind of diverse convolution module, the Re‐parameterization Diversity Convolution (RDC) block, is proposed in the shallow stage of the segmentation network to enrich the feature space. The inference is performed by structural re‐parameterization techniques to decouple the structure at the training and inference stages, thus improving the performance of the model without any inference time cost. In addition, a parallel axial shift MLP (PASMLP) module is proposed, which captures local dependencies by axially shifting the channels of the feature map to capture the information flow from different axes and reduce the network parameters and computational complexity effectively. Extensive experimental results show that the proposed SRP&PASMLP‐Net can provide a 4‐fold improvement in the inference speed and a 17‐fold reduction in model parameters compared with the benchmark network, UNet. It outperforms the state‐of‐the‐art methods on two public datasets, that is, ISIC2018 and PH2, for the SLS.

show abstract

AS-MLP: An Axial Shifted MLP Architecture for Vision

Cited by 42 publications

References 46 publications

PyramidTNT: Improved Transformer-in-Transformer Baselines with Pyramid Architecture

PyramidTNT: Improved Transformer-in-Transformer Baselines with Pyramid Architecture

Are we ready for a new paradigm shift? A Survey on Visual Deep MLP

SRP&PASMLP‐Net: Lightweight skin lesion segmentation network based on structural re‐parameterization and parallel axial shift multilayer perceptron

Contact Info

Product

Resources

About