Papers › X-volution: On the unification of convolution and self-attention

X-volution: On the unification of convolution and self-attention

4 Jun 2021arXiv:2106.02253archive 2025-07-28

Xuanhong Chen, Hang Wang, Bingbing Ni

Convolution and self-attention are acting as two fundamental building blocks in deep neural networks, where the former extracts local image features in a linear way while the latter non-locally encodes high-order contextual relationships. Though essentially complementary to each other, i.e., first-/high-order, stat-of-the-art architectures, i.e., CNNs or transformers lack a principled way to simultaneously apply both operations in a single computational module, due to their heterogeneous computing pattern and excessive burden of global dot-product for visual tasks. In this work, we theoretically derive a global self-attention approximation scheme, which approximates a self-attention via the convolution operation on transformed features. Based on the approximated scheme, we establish a multi-branch elementary module composed of both convolution and self-attention operation, capable of unifying both local and non-local feature interaction. Importantly, once trained, this multi-branch module could be conditionally converted into a single standard convolution operation via structural re-parameterization, rendering a pure convolution styled operator named X-volution, ready to be plugged into any modern networks as an atomic operation. Extensive experiments demonstrate that the proposed X-volution, achieves highly competitive visual understanding improvements (+1.2% top-1 accuracy on ImageNet classification, +1.7 box AP and +1.5 mask AP on COCO detection and segmentation).

PaperPDF

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 ClassificationInstance SegmentationObject Detection

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Classification ImageNet ResNet-50 (X-volution, stage3) Top 1 Accuracy 76.6% #908 of 1060 Archive leaderboard report
Image Classification ImageNet ResNet-34 (X-volution, stage3) Top 1 Accuracy 75% #960 of 1060 Archive leaderboard report
Instance Segmentation COCO minival Mask R-CNN (FPN, X-volution, SA) APL 53.1 #85 of 93 Archive leaderboard report
Instance Segmentation COCO minival Mask R-CNN (FPN, X-volution, SA) APM 40 #85 of 93 Archive leaderboard report
Instance Segmentation COCO minival Mask R-CNN (FPN, X-volution, SA) APS 19.2 #85 of 93 Archive leaderboard report
Instance Segmentation COCO minival Mask R-CNN (FPN, X-volution, SA) mask AP 37.2 #85 of 93 Archive leaderboard report
Object Detection COCO minival Faster R-CNN (FPN, X-volution) AP50 64 #149 of 220 Archive leaderboard report
Object Detection COCO minival Faster R-CNN (FPN, X-volution) AP75 46.4 #149 of 220 Archive leaderboard report
Object Detection COCO minival Faster R-CNN (FPN, X-volution) APL 55 #149 of 220 Archive leaderboard report
Object Detection COCO minival Faster R-CNN (FPN, X-volution) APM 46 #149 of 220 Archive leaderboard report
Object Detection COCO minival Faster R-CNN (FPN, X-volution) APS 26.9 #149 of 220 Archive leaderboard report
Object Detection COCO minival Faster R-CNN (FPN, X-volution) box AP 42.8 #149 of 220 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

Convolution

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