PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
May 29, 2026Pattern Recognition0 citationsOpen Access

A2D2C: Adaptive attention-driven dynamic convolution for local feature adaptation

View Full Paper
TZTianyu ZhangFWFan WanXMXingyu Miao

Key Points

  • To introduce A2D2C, a novel attention-driven dynamic convolution method for local feature adaptation.
  • Utilized multi-point random sampling to efficiently route k base kernels.
  • Implemented A2D2C+ for effective kernel fusion, reducing redundancy in model computations.
  • Demonstrated consistent performance improvements on ImageNet, CIFAR-100, and COCO datasets.

Abstract

• Introduces A 2 D 2 C: attention-driven dynamic convolution for local adaptation. • Uses multi-point random sampling to route and fuse k base kernels efficiently. • Presents A 2 D 2 C ＋ that fuses kernels once, cutting redundancy and MAdds at parity. • Shows consistent gains on ImageNet, CIFAR-100 and COCO with statistical reports.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Zhang et al. (2026) studied this question.

synapsesocial.com/papers/6a192cb4fab5b468c4415892https://doi.org/10.1016/j.patcog.2026.113915
Ask AI
Helpful
Bookmark
Share
View Full Paper