(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network

文章地址:https://arxiv.org/pdf/1909.11856.pdf
作者的项目地址:代码

论文作者:Zheng Hui 西安电子科技大学

一、简单介绍 

受IDN(CVPR2018)的的启发,文章提出了Multi-distillation模块,实现了内存和实时性之间的最优。整个文章的主要的贡献有:

1、提出了轻量化的Multi-distillation网络,来快速和准确的做SR。 

2、提出了自适应裁剪的方法,来实现任意尺度的放大。

3、研究发现网络的深度才是影响速度的主要因素。

参考文献:

Fast and Accurate Single Image Super-Resolution via Information Distillation Network(CVPR2018)

Fast, Accurate, and Lightweight Super-Resolution with Cascading Residual Network(ECCV2018)

加上本文,三篇文章是同一个作者。

 

二、网络结构

 

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第1张图片

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第2张图片

 

class IMDModule(nn.Module):
    def __init__(self, in_channels, distillation_rate=0.25):
        super(IMDModule, self).__init__()
        self.distilled_channels = int(in_channels * distillation_rate)
        self.remaining_channels = int(in_channels - self.distilled_channels)
        self.c1 = conv_layer(in_channels, in_channels, 3)
        self.c2 = conv_layer(self.remaining_channels, in_channels, 3)
        self.c3 = conv_layer(self.remaining_channels, in_channels, 3)
        self.c4 = conv_layer(self.remaining_channels, self.distilled_channels, 3)
        self.act = activation('lrelu', neg_slope=0.05)
        self.c5 = conv_layer(in_channels, in_channels, 1)
        self.cca = CCALayer(self.distilled_channels * 4)

    def forward(self, input):
        out_c1 = self.act(self.c1(input))
        distilled_c1, remaining_c1 = torch.split(out_c1, (self.distilled_channels, self.remaining_channels), dim=1)
        #难道这里就是他所为的distillation模块?其实就是将通道进行分离操作
        out_c2 = self.act(self.c2(remaining_c1))
        distilled_c2, remaining_c2 = torch.split(out_c2, (self.distilled_channels, self.remaining_channels), dim=1)
        out_c3 = self.act(self.c3(remaining_c2))
        distilled_c3, remaining_c3 = torch.split(out_c3, (self.distilled_channels, self.remaining_channels), dim=1)
        out_c4 = self.c4(remaining_c3)
        out = torch.cat([distilled_c1, distilled_c2, distilled_c3, out_c4], dim=1)
        out_fused = self.c5(self.cca(out)) + input
        return out_fused

代码中的torch.spilt就是将通道分离出来。

 

def mean_channels(F):
    assert(F.dim() == 4)
    spatial_sum = F.sum(3, keepdim=True).sum(2, keepdim=True)
    return spatial_sum / (F.size(2) * F.size(3))

def stdv_channels(F):
    assert(F.dim() == 4)
    F_mean = mean_channels(F)
    F_variance = (F - F_mean).pow(2).sum(3, keepdim=True).sum(2, keepdim=True) / (F.size(2) * F.size(3))
    return F_variance.pow(0.5)

# contrast-aware channel attention module
class CCALayer(nn.Module):
    def __init__(self, channel, reduction=16):
        super(CCALayer, self).__init__()

        self.contrast = stdv_channels
        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        self.conv_du = nn.Sequential(
            nn.Conv2d(channel, channel // reduction, 1, padding=0, bias=True),
            nn.ReLU(inplace=True),
            nn.Conv2d(channel // reduction, channel, 1, padding=0, bias=True),
            nn.Sigmoid()
        )


    def forward(self, x):
        y = self.contrast(x) + self.avg_pool(x)
        y = self.conv_du(y)
        return x * y

他这里所为的Contrast-aware channel attention module.实际上就是在SE的global average pooling部分上再加上像素点的均方差。如下图所示,红色的部分为global average pooling,其他部分为像素点对应的均方差。
 

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第3张图片

 

4、实验

训练集:DIV2K dataset。评价指标:PSNR and SSIM on Y channel

网络由6个IMDB模块组成

4.3 模型分析

4.3.1 模型参数

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第4张图片

 

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第5张图片

4.3.2 消融实验

分别验证了CCA(原图3)、PRM(原图2中的灰色部分)、IIC(模块间连接)模块的性能。

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第6张图片

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第7张图片

4.3.3 自适应裁剪模块的验证

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第8张图片

4.4 与现有的最好算法对比

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第9张图片

4.5 运行时间

4.5.1 复杂度分析

4.5.2 运行时间

文章的实验表明,运行时间和网络的深度有关。例如EDSR虽然有43M,但是深度为69,所以运行时间短;而RCAN虽然参数数量较少,但是很深,所以运行慢。

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第10张图片

(IMDN ACMM2019)轻量化Lightweight Image Super-Resolution with Information Multi-distillation Network_第11张图片

method achieves a balance including visual quality, execution speed, and memory consumption.

你可能感兴趣的:(SISR,轻量化SR)