Papers › gSwin: Gated MLP Vision Model with Hierarchical Structure of Shifted Window

gSwin: Gated MLP Vision Model with Hierarchical Structure of Shifted Window

24 Aug 2022arXiv:2208.11718archive 2025-07-28

Mocho Go, Hideyuki Tachibana

Following the success in language domain, the self-attention mechanism (transformer) is adopted in the vision domain and achieving great success recently. Additionally, as another stream, multi-layer perceptron (MLP) is also explored in the vision domain. These architectures, other than traditional CNNs, have been attracting attention recently, and many methods have been proposed. As one that combines parameter efficiency and performance with locality and hierarchy in image recognition, we propose gSwin, which merges the two streams; Swin Transformer and (multi-head) gMLP. We showed that our gSwin can achieve better accuracy on three vision tasks, image classification, object detection and semantic segmentation, than Swin Transformer, with smaller model size.

PaperPDF

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 ClassificationInstance SegmentationObject DetectionSemantic Segmentationimage-classificationobject-detection

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Classification ImageNet gSwin-S GFLOPs 7.0 #472 of 1060 Archive leaderboard report
Image Classification ImageNet gSwin-S Number of params 39.8M #472 of 1060 Archive leaderboard report
Image Classification ImageNet gSwin-S Top 1 Accuracy 83.01% #472 of 1060 Archive leaderboard report
Image Classification ImageNet gSwin-T GFLOPs 3.6 #613 of 1060 Archive leaderboard report
Image Classification ImageNet gSwin-T Number of params 21.8M #613 of 1060 Archive leaderboard report
Image Classification ImageNet gSwin-T Top 1 Accuracy 81.71% #613 of 1060 Archive leaderboard report
Image Classification ImageNet gSwin-VT GFLOPs 2.3 #706 of 1060 Archive leaderboard report
Image Classification ImageNet gSwin-VT Number of params 15.5M #706 of 1060 Archive leaderboard report
Image Classification ImageNet gSwin-VT Top 1 Accuracy 80.32% #706 of 1060 Archive leaderboard report
Instance Segmentation COCO test-dev gSwin-S mask AP 45.03 #44 of 112 Archive leaderboard report
Instance Segmentation COCO test-dev gSwin-T mask AP 44.16 #46 of 112 Archive leaderboard report
Instance Segmentation COCO test-dev gSwin-VT mask AP 42.87 #52 of 112 Archive leaderboard report
Semantic Segmentation ADE20K val gSwin-S Pixel Accuracy 83.43 #55 of 95 Archive leaderboard report
Semantic Segmentation ADE20K val gSwin-S mIoU 49.69 #55 of 95 Archive leaderboard report
Semantic Segmentation ADE20K val gSwin-T Pixel Accuracy 82.60 #64 of 95 Archive leaderboard report
Semantic Segmentation ADE20K val gSwin-T mIoU 47.63 #64 of 95 Archive leaderboard report
Semantic Segmentation ADE20K val gSwin-VT Pixel Accuracy 81.79 #83 of 95 Archive leaderboard report
Semantic Segmentation ADE20K val gSwin-VT mIoU 45.07 #83 of 95 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

Absolute Position EncodingsAdamAttentionBPEDense ConnectionsDropoutLabel SmoothingLayer NormalizationLinear LayerMulti-Head AttentionPosition-Wise Feed-Forward LayerResidual ConnectionSoftmaxSpatial Gating UnitStochastic DepthSwin TransformerTransformergMLP

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