[论文解读] Lightweight Image Super-Resolution with Adaptive Weighted Learning Network
提出一个轻量级的 SISR 网络 (AWSRN),具有局部融合块和自适应加权多尺度模块,在参数量和计算量相近的情况下实现卓越性能。
Deep learning has been successfully applied to the single-image super-resolution (SISR) task with great performance in recent years. However, most convolutional neural network based SR models require heavy computation, which limit their real-world applications. In this work, a lightweight SR network, named Adaptive Weighted Super-Resolution Network (AWSRN), is proposed for SISR to address this issue. A novel local fusion block (LFB) is designed in AWSRN for efficient residual learning, which consists of stacked adaptive weighted residual units (AWRU) and a local residual fusion unit (LRFU). Moreover, an adaptive weighted multi-scale (AWMS) module is proposed to make full use of features in reconstruction layer. AWMS consists of several different scale convolutions, and the redundancy scale branch can be removed according to the contribution of adaptive weights in AWMS for lightweight network. The experimental results on the commonly used datasets show that the proposed lightweight AWSRN achieves superior performance on x2, x3, x4, and x8 scale factors to state-of-the-art methods with similar parameters and computational overhead. Code is avaliable at: https://github.com/ChaofWang/AWSRN
研究动机与目标
- 降低单图像超分辨率的计算成本。
- 设计具有有效残差学习的轻量级架构。
- 在重建中利用自适应加权来选择有用的尺度和特征。
提出的方法
- 引入一个轻量级的 Adaptive Weighted Super-Resolution Network (AWSRN)。
- 开发一个局部融合块 (LFB),由堆叠的自适应加权残差单元 (AWRU) 和局部残差融合单元 (LRFU) 构成。
- 提出一个自适应加权多尺度 (AWMS) 模块,具备多尺度卷积和在自适应权重引导下可移除的冗余分支。
- 在重建中利用自适应加权残差学习以实现高效特征融合。
- 在标准 SR 基准测试上评估在 x2、x3、x4 和 x8 比例下的性能。
实验结果
研究问题
- RQ1在比重量级模型更轻的网络中,结合自适应加权,是否能达到与之竞争的 SR 质量?
- RQ2局部融合块 (LFB) 如何在紧凑架构中提升残差学习的效率?
- RQ3自适应加权多尺度 (AWMS) 模块对重建质量和模型效率有何影响?
- RQ4AWRSN 是否在多个放大因子(x2、x3、x4、x8)下,在相似的参数数量和计算开销下保持卓越性能?
主要发现
- AWSRN 在参数和计算开销相近的情况下实现优于现有方法的性能。
- 局部融合块通过堆叠的 AWRUs 和 LRFU 实现高效的残差学习。
- AWMS 利用多尺度卷积和自适应权重来利用有信息的尺度,同时移除冗余分支。
- 该方法在多个放大因子(x2、x3、x4、x8)下都有效。
- 结果在论文所述的常用 SR 数据集上得到展示。
更好的研究,从现在开始
从论文设计到论文写作,大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成,并经人工编辑审核。