import paddleimport paddle.nn as nnfrom math import ceil登录后复制 ? ?

基础模块

• 包含 ConvBN、Swish 和 SE 模块

def ConvBN(out, in_channels, channels, kernel=1, stride=1, pad=0, num_group=1, act=None): out.append(nn.Conv2D(in_channels, channels, kernel, stride, pad, groups=num_group, bias_attr=False)) out.append(nn.BatchNorm2D(channels)) if act == 'swish': out.append(Swish()) elif act == 'relu': out.append(nn.ReLU()) elif act == 'relu6': out.append(nn.ReLU6())class Swish(nn.Layer): def __init__(self): super(Swish, self).__init__() def forward(self, x): return x * nn.functional.sigmoid(x)class SE(nn.Layer): def __init__(self, in_channels, channels, se_ratio=12): super(SE, self).__init__() self.avg_pool = nn.AdaptiveAvgPool2D(1) self.fc = nn.Sequential( nn.Conv2D(in_channels, channels // se_ratio, kernel_size=1, padding=0), nn.BatchNorm2D(channels // se_ratio), nn.ReLU(), nn.Conv2D(channels // se_ratio, channels, kernel_size=1, padding=0), nn.Sigmoid() ) def forward(self, x): y = self.avg_pool(x) y = self.fc(y) return x * y登录后复制 ? ?

LinearBottleneck

class LinearBottleneck(nn.Layer): def __init__(self, in_channels, channels, t, stride, use_se=True, se_ratio=12): super(LinearBottleneck, self).__init__() self.use_shortcut = stride == 1 and in_channels <= channels self.in_channels = in_channels self.out_channels = channels out = [] if t != 1: dw_channels = in_channels * t ConvBN(out, in_channels=in_channels, channels=dw_channels, act='swish') else: dw_channels = in_channels ConvBN(out, in_channels=dw_channels, channels=dw_channels, kernel=3, stride=stride, pad=1, num_group=dw_channels) if use_se: out.append(SE(dw_channels, dw_channels, se_ratio)) out.append(nn.ReLU6()) ConvBN(out, in_channels=dw_channels, channels=channels) self.out = nn.Sequential(*out) def forward(self, x): out = self.out(x) if self.use_shortcut: out[:, 0:self.in_channels] += x return out登录后复制 ? ?

ReXNet

• 模型主体

class ReXNetV1(nn.Layer): def __init__(self, input_ch=16, final_ch=180, width_mult=1.0, depth_mult=1.0, class_dim=1000, use_se=True, se_ratio=12, dropout_ratio=0.2): super(ReXNetV1, self).__init__() layers = [1, 2, 2, 3, 3, 5] strides = [1, 2, 2, 2, 1, 2] use_ses = [False, False, True, True, True, True] layers = [ceil(element * depth_mult) for element in layers] strides = sum([[element] + [1] * (layers[idx] - 1) for idx, element in enumerate(strides)], []) if use_se: use_ses = sum([[element] * layers[idx] for idx, element in enumerate(use_ses)], []) else: use_ses = [False] * sum(layers[:]) ts = [1] * layers[0] + [6] * sum(layers[1:]) self.depth = sum(layers[:]) * 3 stem_channel = 32 / width_mult if width_mult < 1.0 else 32 inplanes = input_ch / width_mult if width_mult < 1.0 else input_ch features = [] in_channels_group = [] channels_group = [] # The following channel configuration is a simple instance to make each layer become an expand layer. for i in range(self.depth // 3): if i == 0: in_channels_group.append(int(round(stem_channel * width_mult))) channels_group.append(int(round(inplanes * width_mult))) else: in_channels_group.append(int(round(inplanes * width_mult))) inplanes += final_ch / (self.depth // 3 * 1.0) channels_group.append(int(round(inplanes * width_mult))) ConvBN(features, 3, int(round(stem_channel * width_mult)), kernel=3, stride=2, pad=1, act='swish') for block_idx, (in_c, c, t, s, se) in enumerate(zip(in_channels_group, channels_group, ts, strides, use_ses)): features.append(LinearBottleneck(in_channels=in_c, channels=c, t=t, stride=s, use_se=se, se_ratio=se_ratio)) pen_channels = int(1280 * width_mult) ConvBN(features, c, pen_channels, act='swish') features.append(nn.AdaptiveAvgPool2D(1)) self.features = nn.Sequential(*features) self.output = nn.Sequential( nn.Dropout(dropout_ratio), nn.Conv2D(pen_channels, class_dim, 1)) def forward(self, x): x = self.features(x) x = self.output(x).squeeze() return x登录后复制 ? ?

预设模型

def rexnet_100(pretrained=False, **kwargs): model = ReXNetV1(width_mult=1.0, **kwargs) return modeldef rexnet_130(pretrained=False, **kwargs): model = ReXNetV1(width_mult=1.3, **kwargs) return modeldef rexnet_150(pretrained=False, **kwargs): model = ReXNetV1(width_mult=1.5, **kwargs) return modeldef rexnet_200(pretrained=False, **kwargs): model = ReXNetV1(width_mult=2.0, **kwargs) return modeldef rexnet_300(pretrained=False, **kwargs): model = ReXNetV1(width_mult=3.0, **kwargs) return model登录后复制 ? ?

模型验证

• 加载转换后的官方预训练模型，使用 ImageNet 1k 验证集进行模型测试
• 验证精度是否与官方标称一致

解压数据集

!mkdir ~/data/ILSVRC2012!tar -xf ~/data/data68594/ILSVRC2012_img_val.tar -C ~/data/ILSVRC2012登录后复制 ? ?

构建数据集

• 构建一个 ILSVRC2012 ImageNet 1k 的数据集

import osimport paddleimport numpy as npfrom PIL import Imageclass ILSVRC2012(paddle.io.Dataset): def __init__(self, root, label_list, transform): self.transform = transform self.root = root self.label_list = label_list self.load_datas() def load_datas(self): self.imgs = [] self.labels = [] with open(self.label_list, 'r') as f: for line in f: img, label = line[:-1].split(' ') self.imgs.append(os.path.join(self.root, img)) self.labels.append(int(label)) def __getitem__(self, idx): label = self.labels[idx] image = self.imgs[idx] image = Image.open(image).convert('RGB') image = self.transform(image) return image.astype('float32'), np.array(label).astype('int64') def __len__(self): return len(self.imgs)登录后复制 ? ?

模型验证

• ReXNet 的官方标称精度如下：