使用PyTorch實作ResNet並提取指定層輸出之特徵

最近在oodformer的程式碼中看到一個不錯的ResNet神經網路實作。

在ResNet這個class中，除了有一般的模型forward外，也加入了以下功能：

(1)提取經過多層的特徵(使用list保存，可進行indexing)
def feature_list(self, x):
(2)提取經過N層後的特徵
def intermediate_forward(self, x, layer_index):
(3)提取倒數第二層特徵(常拿來來視覺化t-SNE)
def penultimate_forward(self, x):

本篇文章記錄其作法：

# function to extract the multiple features
def feature_list(self, x):
    out_list = []
    out = F.relu(self.bn1(self.conv1(x)))
    out_list.append(out)
    out = self.layer1(out)
    out_list.append(out)
    out = self.layer2(out)
    out_list.append(out)
    out = self.layer3(out)
    out_list.append(out)
    out = self.layer4(out)
    out_list.append(out)
    out = F.avg_pool2d(out, 4)
    out = out.view(out.size(0), -1)
    y = self.linear(out)
    return y, out_list
# function to extract a specific feature
def intermediate_forward(self, x, layer_index):
    out = F.relu(self.bn1(self.conv1(x)))
    if layer_index == 1:
        out = self.layer1(out)
    elif layer_index == 2:
        out = self.layer1(out)
        out = self.layer2(out)
    elif layer_index == 3:
        out = self.layer1(out)
        out = self.layer2(out)
        out = self.layer3(out)
    elif layer_index == 4:
        out = self.layer1(out)
        out = self.layer2(out)
        out = self.layer3(out)
        out = self.layer4(out)               
    return out
# function to extract the penultimate features
def penultimate_forward(self, x):
    out = F.relu(self.bn1(self.conv1(x)))
    out = self.layer1(out)
    out = self.layer2(out)
    out = self.layer3(out)
    penultimate = self.layer4(out)
    out = F.avg_pool2d(penultimate, 4)
    out = out.view(out.size(0), -1)
    y = self.linear(out)
    return y, penultimate

程式碼來源

oodformer/resnet.py at master · rajatkoner08/oodformer