Papers › UniNet: Unified Architecture Search with Convolution, Transformer, and MLP

UniNet: Unified Architecture Search with Convolution, Transformer, and MLP

8 Oct 2021arXiv:2110.04035archive 2025-07-28

Jihao Liu, Hongsheng Li, Guanglu Song, Xin Huang, Yu Liu

Recently, transformer and multi-layer perceptron (MLP) architectures have achieved impressive results on various vision tasks. A few works investigated manually combining those operators to design visual network architectures, and can achieve satisfactory performances to some extent. In this paper, we propose to jointly search the optimal combination of convolution, transformer, and MLP for building a series of all-operator network architectures with high performances on visual tasks. We empirically identify that the widely-used strided convolution or pooling based down-sampling modules become the performance bottlenecks when the operators are combined to form a network. To better tackle the global context captured by the transformer and MLP operators, we propose two novel context-aware down-sampling modules, which can better adapt to the global information encoded by transformer and MLP operators. To this end, we jointly search all operators and down-sampling modules in a unified search space. Notably, Our searched network UniNet (Unified Network) outperforms state-of-the-art pure convolution-based architecture, EfficientNet, and pure transformer-based architecture, Swin-Transformer, on multiple public visual benchmarks, ImageNet classification, COCO object detection, and ADE20K semantic segmentation.

PaperPDFConference PDF

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image ClassificationObject DetectionSemantic Segmentationobject-detection

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Classification ImageNet UniNet-B5 GFLOPs 23.2 #250 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B5 Number of params 73.5M #250 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B5 Top 1 Accuracy 85.2% #250 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B4 GFLOPs 9.9 #339 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B4 Number of params 73.5M #339 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B4 Top 1 Accuracy 84.2% #339 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B2 GFLOPs 2.4 #510 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B2 Number of params 22.5M #510 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B2 Top 1 Accuracy 82.7% #510 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B1 GFLOPs 0.99 #703 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B1 Number of params 14M #703 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B1 Top 1 Accuracy 80.4% #703 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B0 GFLOPs 0.56 #783 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B0 Number of params 11.9M #783 of 1060 Archive leaderboard report
Image Classification ImageNet UniNet-B0 Top 1 Accuracy 79.1% #783 of 1060 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

1x1 ConvolutionAverage PoolingBatch NormalizationConvolutionDense ConnectionsDepthwise ConvolutionDepthwise Separable ConvolutionDropoutInverted Residual BlockPointwise ConvolutionRMSPropReLUSigmoid ActivationSqueeze-and-Excitation Block

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections