Attention in Attention Network for Image Super-Resolution
arXiv:2104.09497
Abstract
Convolutional neural networks have allowed remarkable advances in single image super-resolution (SISR) over the last decade. Among recent advances in SISR, attention mechanisms are crucial for high-performance SR models. However, the attention mechanism remains unclear on why and how it works in SISR. In this work, we attempt to quantify and visualize attention mechanisms in SISR and show that not all attention modules are equally beneficial. We then propose attention in attention network (AN) for more efficient and accurate SISR. Specifically, AN consists of a non-attention branch and a coupling attention branch. A dynamic attention module is proposed to generate weights for these two branches to suppress unwanted attention adjustments dynamically, where the weights change adaptively according to the input features. This allows attention modules to specialize to beneficial examples without otherwise penalties and thus greatly improve the capacity of the attention network with few parameters overhead. Experimental results demonstrate that our final model AN could achieve superior trade-off performances comparing with state-of-the-art networks of similar sizes. Codes are available at https://github.com/haoyuc/A2N.
arXiv comment11 pages, 10 figures. Codes are available at https://github.com/haoyuc/A2N
References in corpus (8)
- Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
- Sketch-based Manga Retrieval using Manga109 Dataset
- ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks
- Residual Non-local Attention Networks for Image Restoration
- Lightweight Image Super-Resolution with Adaptive Weighted Learning Network
- Efficient Image Super-Resolution Using Pixel Attention
- MAMNet: Multi-path Adaptive Modulation Network for Image Super-Resolution
- Hierarchical Residual Attention Network for Single Image Super-Resolution