【第五期论文复现赛-语义分割】DMNet

class DCM(nn.Layer):
    """
    Dynamic Convolutional Module used in DMNet.

    Args:
        filter_size (int): The filter size of generated convolution kernel used in Dynamic Convolutional Module.
        fusion (bool): Add one conv to fuse DCM output feature.
        in_channels (int): Input channels.
        channels (int): Channels after modules, before conv_seg.
    """
    def __init__(self, filter_size, fusion, in_channels, channels):
        super().__init__()
        self.filter_size = filter_size          # 滤波器尺寸
        self.fusion = fusion                    # 是否在DCM的输出特征图后面加上一个卷积层
        self.channels = channels                # 输出通道数
        
        pad = (self.filter_size - 1) // 2       # 计算pad
        if (self.filter_size - 1) % 2 == 0:
            self.pad = (pad, pad, pad, pad)        else:
            self.pad = (pad + 1, pad, pad + 1, pad)

        self.avg_pool = nn.AdaptiveAvgPool2D(filter_size)  # 通过自适应池化产生filter的权重
        self.filter_gen_conv = nn.Conv2D(in_channels, channels, 1)
        self.input_redu_conv = layers.ConvBNReLU(in_channels, channels, 1)  # 1x1的卷积减少特征图通道

        self.norm = layers.SyncBatchNorm(channels)
        self.act = nn.ReLU()        if self.fusion:
            self.fusion_conv = layers.ConvBNReLU(channels, channels, 1)    def forward(self, x):
        generated_filter = self.filter_gen_conv(
            self.avg_pool(x))        # 分支二：生成滤波器权重
        x = self.input_redu_conv(x)  # 分支一：减少特征图通道
        b, c, h, w = x.shape
        x = x.reshape([1, b * c, h, w])
        generated_filter = generated_filter.reshape(
            [b * c, 1, self.filter_size, self.filter_size])
        
        x = F.pad(x, self.pad, mode='constant', value=0)  
        output = F.conv2d(x, weight=generated_filter, groups=b * c)  # 深度可分离卷积
        output = output.reshape([b, self.channels, h, w])
        output = self.norm(output)
        output = self.act(output)        if self.fusion:
            output = self.fusion_conv(output)        return output

# step 1: 解压数据%cd ~/data/data64550
!tar -xf cityscapes.tar
%cd ~/

# step 2: 模型训练%cd ~/DMNet_paddle/
!python train.py --config configs/dmnet/dmnet_cityscapes_1024x512_100k.yml --num_workers 4 --do_eval --use_vdl --log_iter 20 --save_interval 2000

# step 3: 模型验证--此处给出了在Tesla V100 * 4环境下训练出的模型权重%cd ~/DMNet_paddle/
!python val.py --config configs/dmnet/dmnet_cityscapes_1024x512_100k.yml --model_path output/best_model/model.pdparams