jittor.models.res2net 源代码

import jittor as jt
from jittor import nn
from jittor import Module
from jittor import init
from jittor.contrib import concat, argmax_pool
import math


model_urls = {
    'res2net50_14w_8s': 'jittorhub://res2net50_14w_8s.pkl',
    'res2net50_26w_4s': 'jittorhub://res2net50_26w_4s.pkl',
    'res2net50_26w_6s': 'jittorhub://res2net50_26w_6s.pkl',
    'res2net50_26w_8s': 'jittorhub://res2net50_26w_8s.pkl',
    'res2net50_48w_2s': 'jittorhub://res2net50_48w_2s.pkl',
    'res2net101_26w_4s': 'jittorhub://res2net101_26w_4s.pkl',
}


class Bottle2neck(nn.Module):
    expansion = 4

    def __init__(self, inplanes, planes, stride=1, downsample=None, baseWidth=26, scale = 4, stype='normal'):
        super(Bottle2neck, self).__init__()

        width = int(math.floor(planes * (baseWidth/64.0)))
        self.conv1 = nn.Conv2d(inplanes, width*scale, kernel_size=1, bias=False)
        self.bn1 = nn.BatchNorm2d(width*scale)
        
        if scale == 1:
          self.nums = 1
        else:
          self.nums = scale -1
        if stype == 'stage':
            self.pool = nn.AvgPool2d(kernel_size=3, stride = stride, padding=1)
        convs = []
        bns = []
        for i in range(self.nums):
          convs.append(nn.Conv2d(width, width, kernel_size=3, stride = stride, padding=1, bias=False))
          bns.append(nn.BatchNorm2d(width))
        self.convs = nn.ModuleList(convs)
        self.bns = nn.ModuleList(bns)

        self.conv3 = nn.Conv2d(width*scale, planes * self.expansion, kernel_size=1, bias=False)
        self.bn3 = nn.BatchNorm2d(planes * self.expansion)

        self.relu = nn.ReLU()
        self.downsample = downsample
        self.stype = stype
        self.scale = scale
        self.width  = width

    def execute(self, x):
        residual = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        spx = jt.split(out, self.width, 1)
        for i in range(self.nums):
          if i==0 or self.stype=='stage':
            sp = spx[i]
          else:
            sp = sp + spx[i]
          sp = self.convs[i](sp)
          sp = self.relu(self.bns[i](sp))
          if i==0:
            out = sp
          else:
            out = jt.concat((out, sp), 1)
        if self.scale != 1 and self.stype=='normal':
          out = jt.concat((out, spx[self.nums]),1)
        elif self.scale != 1 and self.stype=='stage':
          out = jt.concat((out, self.pool(spx[self.nums])),1)

        out = self.conv3(out)
        out = self.bn3(out)

        if self.downsample is not None:
            residual = self.downsample(x)

        out += residual
        out = self.relu(out)

        return out

class Res2Net(nn.Module):
    '''
    Res2Net源自论文 `Res2Net: A New Multi-scale Backbone Architecture <https://arxiv.org/abs/1904.01169>`__, Res2Net通过在ResNet的基础上引入一个新的构建块, 该构建块将输入的特征图分成几个子集, 并对每个子集应用不同的滤波器, 最后将这些子集重组来形成最终的输出。这种方法能够以较小的额外计算复杂度显著地提高模型的表达能力。

        参数: 
            - block (nn.Module): 构建块类型, 该类型需要具备分支扩展能力, 例如Res2Net中的BottleNeck等。
            - layers (List[int]): 一个包含四个整数的列表, 指定每个层的构建块的数量。
            - baseWidth (int, optional): block中基本宽度, 决定了网络的组宽。默认值: 26
            - scale (int, optional): 每个block中的分支数(尺度)。默认值: 4
            - num_classes (int, optional): 分类的类别数。默认值: 1000

        属性:
            - conv1 (nn.Conv2d): 输入层的卷积层。
            - bn1 (nn.BatchNorm2d): 输入层的批归一化层。
            - relu (nn.ReLU): 激活函数。
            - maxpool (nn.MaxPool2d): 最大池化层。
            - layer1 (nn.Sequential): 第一个layer。
            - layer2 (nn.Sequential): 第二个layer。
            - layer3 (nn.Sequential): 第三个layer。
            - layer4 (nn.Sequential): 第四个layer。
            - avgpool (nn.AdaptiveAvgPool2d): 自适应平均池化层。
            - fc (nn.Linear): 全连接层。
        
        代码示例:  
            >>> import jittor as jt
            >>> from jittor.models.res2net import Res2Net, Bottle2neck
            >>> model = Res2Net(Bottle2neck, [3, 4, 6, 3])
            >>> input = jt.randn(4, 3, 224, 224) # 假设是一个形状为[批次大小, 通道数, 高度, 宽度]的张量
            >>> model(input).shape
            [4,1000,]

    '''
    def __init__(self, block, layers, baseWidth = 26, scale = 4, num_classes=1000):
        self.inplanes = 64
        super(Res2Net, self).__init__()
        self.baseWidth = baseWidth
        self.scale = scale
        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,
                               bias=False)
        self.bn1 = nn.BatchNorm2d(64)
        self.relu = nn.ReLU()
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        self.layer1 = self._make_layer(block, 64, layers[0])
        self.layer2 = self._make_layer(block, 128, layers[1], stride=2)
        self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
        self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
        self.avgpool = nn.AdaptiveAvgPool2d(1)
        self.fc = nn.Linear(512 * block.expansion, num_classes)

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, nn.BatchNorm2d):
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)

    def _make_layer(self, block, planes, blocks, stride=1):
        downsample = None
        if stride != 1 or self.inplanes != planes * block.expansion:
            downsample = nn.Sequential(
                nn.Conv2d(self.inplanes, planes * block.expansion,
                          kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(planes * block.expansion),
            )

        layers = []
        layers.append(block(self.inplanes, planes, stride, downsample=downsample, 
                        stype='stage', baseWidth = self.baseWidth, scale=self.scale))
        self.inplanes = planes * block.expansion
        for i in range(1, blocks):
            layers.append(block(self.inplanes, planes, baseWidth = self.baseWidth, scale=self.scale))

        return nn.Sequential(*layers)

    def execute(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.maxpool(x)

        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)

        x = self.avgpool(x)
        x = x.view(x.size(0), -1)
        x = self.fc(x)

        return x


[文档] def res2net50(pretrained=False, **kwargs): ''' 构建一个Res2Net50模型 Res2Net50源自论文 `Res2Net: A New Multi-scale Backbone Architecture <https://arxiv.org/abs/1904.01169>`__, 是ResNet模型的一种变体。 参数: - `pretrained` (bool, optional): 如果为 `True`, 返回一个在ImageNet上预训练的模型。默认为 `False`。 - `**kwargs` (dict, optional): 用于传递额外的关键字参数。 返回值: - 返回构建好的Res2Net模型实例。如果 `pretrained` 为 `True`, 则返回在ImageNet上预训练的模型。 代码示例: >>> import jittor as jt >>> from jittor.models.res2net import * >>> net = res2net50(pretrained=True) >>> x = jt.rand(1, 3, 224, 224) >>> y = net(x) >>> y.shape [1, 1000] ''' model = Res2Net(Bottle2neck, [3, 4, 6, 3], baseWidth = 26, scale = 4, **kwargs) if pretrained: model.load(model_urls['res2net50_26w_4s']) return model
def res2net50_26w_4s(pretrained=False, **kwargs): ''' 构建一个Res2Net50_26w_4s模型 Res2Net50_26w_4s源自论文 `Res2Net: A New Multi-scale Backbone Architecture <https://arxiv.org/abs/1904.01169>`__, 是ResNet模型的一种变体。 参数: - `pretrained` (bool, optional): 如果为 `True`, 返回一个在ImageNet上预训练的模型。默认为 `False`。 - `**kwargs` (dict, optional): 用于传递额外的关键字参数。 返回值: - 返回构建好的Res2Net_26w_4s模型实例。如果 `pretrained` 为 `True`, 则返回在ImageNet上预训练的模型。 代码示例: >>> import jittor as jt >>> from jittor.models.res2net import * >>> net = res2net50_26w_4s(pretrained=True) >>> x = jt.rand(1, 3, 224, 224) >>> y = net(x) >>> y.shape [1, 1000] ''' model = Res2Net(Bottle2neck, [3, 4, 6, 3], baseWidth = 26, scale = 4, **kwargs) if pretrained: model.load(model_urls['res2net50_26w_4s']) return model def res2net101_26w_4s(pretrained=False, **kwargs): ''' 构建一个Res2Net101_26w_4s模型 Res2Net101_26w_4s源自论文 `Res2Net: A New Multi-scale Backbone Architecture <https://arxiv.org/abs/1904.01169>`__, 是ResNet模型的一种变体。 参数: - `pretrained` (bool, optional): 如果为 `True`, 返回一个在ImageNet上预训练的模型。默认为 `False`。 - `**kwargs` (dict, optional): 用于传递额外的关键字参数。 返回值: - 返回构建好的Res2Net101_26w_4s模型实例。如果 `pretrained` 为 `True`, 则返回在ImageNet上预训练的模型。 代码示例: >>> import jittor as jt >>> from jittor.models.res2net import * >>> net = res2net101_26w_4s(pretrained=True) >>> x = jt.rand(1, 3, 224, 224) >>> y = net(x) >>> y.shape [1, 1000] ''' model = Res2Net(Bottle2neck, [3, 4, 23, 3], baseWidth = 26, scale = 4, **kwargs) if pretrained: model.load(model_urls['res2net101_26w_4s']) return model res2net101 = res2net101_26w_4s def res2net50_26w_6s(pretrained=False, **kwargs): ''' 构建一个Res2Net50_26w_6s模型 Res2Net50_26w_6s源自论文 `Res2Net: A New Multi-scale Backbone Architecture <https://arxiv.org/abs/1904.01169>`__, 是ResNet模型的一种变体。 参数: - `pretrained` (bool, optional): 如果为 `True`, 返回一个在ImageNet上预训练的模型。默认为 `False`。 - `**kwargs` (dict, optional): 用于传递额外的关键字参数。 返回值: - 返回构建好的Res2Net50_26w_6s模型实例。如果 `pretrained` 为 `True`, 则返回在ImageNet上预训练的模型。 代码示例: >>> import jittor as jt >>> from jittor.models.res2net import * >>> net = res2net50_26w_6s(pretrained=True) >>> x = jt.rand(1, 3, 224, 224) >>> y = net(x) >>> y.shape [1, 1000] ''' model = Res2Net(Bottle2neck, [3, 4, 6, 3], baseWidth = 26, scale = 6, **kwargs) if pretrained: model.load(model_urls['res2net50_26w_6s']) return model def res2net50_26w_8s(pretrained=False, **kwargs): ''' 构建一个Res2Net50_26w_8s模型 Res2Net50_26w_8s源自论文 `Res2Net: A New Multi-scale Backbone Architecture <https://arxiv.org/abs/1904.01169>`__, 是ResNet模型的一种变体。 参数: - `pretrained` (bool, optional): 如果为 `True`, 返回一个在ImageNet上预训练的模型。默认为 `False`。 - `**kwargs` (dict, optional): 用于传递额外的关键字参数。 返回值: - 返回构建好的Res2Net50_26w_8s模型实例。如果 `pretrained` 为 `True`, 则返回在ImageNet上预训练的模型。 代码示例: >>> import jittor as jt >>> from jittor.models.res2net import * >>> net = res2net50_26w_8s(pretrained=True) >>> x = jt.rand(1, 3, 224, 224) >>> y = net(x) >>> y.shape [1, 1000] ''' model = Res2Net(Bottle2neck, [3, 4, 6, 3], baseWidth = 26, scale = 8, **kwargs) if pretrained: model.load(model_urls['res2net50_26w_8s']) return model def res2net50_48w_2s(pretrained=False, **kwargs): ''' 构建一个Res2Net50_48w_2s模型 Res2Net50_48w_2s源自论文 `Res2Net: A New Multi-scale Backbone Architecture <https://arxiv.org/abs/1904.01169>`__, 是ResNet模型的一种变体。 参数: - `pretrained` (bool, optional): 如果为 `True`, 返回一个在ImageNet上预训练的模型。默认为 `False`。 - `**kwargs` (dict, optional): 用于传递额外的关键字参数。 返回值: - 返回构建好的Res2Net50_48w_2s模型实例。如果 `pretrained` 为 `True`, 则返回在ImageNet上预训练的模型。 代码示例: >>> import jittor as jt >>> from jittor.models.res2net import * >>> net = res2net50_48w_2s(pretrained=True) >>> x = jt.rand(1, 3, 224, 224) >>> y = net(x) >>> y.shape [1, 1000] ''' model = Res2Net(Bottle2neck, [3, 4, 6, 3], baseWidth = 48, scale = 2, **kwargs) if pretrained: model.load(model_urls['res2net50_48w_2s']) return model def res2net50_14w_8s(pretrained=False, **kwargs): ''' 构建一个Res2Net50_14w_8s模型 Res2Net50_14w_8s源自论文 `Res2Net: A New Multi-scale Backbone Architecture <https://arxiv.org/abs/1904.01169>`__, 是ResNet模型的一种变体。 参数: - `pretrained` (bool, optional): 如果为 `True`, 返回一个在ImageNet上预训练的模型。默认为 `False`。 - `**kwargs` (dict, optional): 用于传递额外的关键字参数。 返回值: - 返回构建好的Res2Net50_14w_8s模型实例。如果 `pretrained` 为 `True`, 则返回在ImageNet上预训练的模型。 代码示例: >>> import jittor as jt >>> from jittor.models.res2net import * >>> net = res2net50_14w_8s(pretrained=True) >>> x = jt.rand(1, 3, 224, 224) >>> y = net(x) >>> y.shape [1, 1000] ''' model = Res2Net(Bottle2neck, [3, 4, 6, 3], baseWidth = 14, scale = 8, **kwargs) if pretrained: model.load(model_urls['res2net50_14w_8s']) return model