Conditional Convolution Residual Network for Efficient Super-Resolution

With the continuous development of deep learning, single-image super-resolution (SISR) based on convolutional neural networks (CNNs) has made significant progress. Although CNN-based methods have achieved great success, these methods are difficult to apply to edge devices due to the need for large amounts of computing resources. To address this problem, the latest advancements in efficient SISR techniques focus on reducing the number of parameters and multiply-add operations (MAdds). In this paper, we propose a novel Conditional Convolution Residual Network (CCRN) to tackle this challenge. The main idea is to use conditional convolution instead of ordinary convolutional layers for residual feature learning and to combine Contrast-aware Channel Attention (CCA) and Enhanced Spatial Attention (ESA) mechanisms to improve the model's performance. The model's performance is ensured while reducing the computational complexity. Experimental results demonstrate that CCRN has fewer MAdds than existing SISR methods while achieving state-of-the-art performance.