Dimension-aware attention for efficient mobile networks
作者：
Highlights：
• An attention mechanism (the DAA block) is developed for feature enhancement.
• Multi-branch factorization design enables low redundancy and high efficiency.
• The DAA block introduces a small computational cost with large receptive fields.
• The DAA block can be easily integrated with existing mobile networks.
• Experiments on multiple vision tasks show the effectiveness of our method.
摘要
•An attention mechanism (the DAA block) is developed for feature enhancement.•Multi-branch factorization design enables low redundancy and high efficiency.•The DAA block introduces a small computational cost with large receptive fields.•The DAA block can be easily integrated with existing mobile networks.•Experiments on multiple vision tasks show the effectiveness of our method.
论文关键词：Efficient mobile networks,Attention mechanism,Feature enhancement,Multi-branch factorization,Multi-dimensional information
论文评审过程：Received 6 May 2022, Revised 5 July 2022, Accepted 13 July 2022, Available online 16 July 2022, Version of Record 21 July 2022.
论文官网地址：https://doi.org/10.1016/j.patcog.2022.108899
