为什么ResNet深度残差网络广受好评呢?理论部分参考
一、数据集下载
CIFAR-10 and CIFAR-100 datasets (toronto.edu)
二、预训练权重下载
'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth', 'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth', 'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth', 'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth', 'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',
三、Demo介绍
构造一个resnet-20,在cifar-10上参考论文中的设置进行训练,达到论文给出的精度。
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
if __name__ == "__main__":
# config
train_dir = os.path.join("..", "data", "cifar-10", "cifar10_train")
test_dir = os.path.join("..", "data", "cifar-10", "cifar10_test")
now_time = datetime.now()
time_str = datetime.strftime(now_time, '%m-%d_%H-%M')
log_dir = os.path.join(BASE_DIR, "..", "results", time_str)
if not os.path.exists(log_dir):
os.makedirs(log_dir)
class_names = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
num_classes = 10
MAX_EPOCH = 182 # 182 # 64000 / (45000 / 128) = 182 epochs
BATCH_SIZE = 128
LR = 0.1
log_interval = 1
val_interval = 1
start_epoch = -1
milestones = [92, 136] # divide it by 10 at 32k and 48k iterations
# ============================ step 1/5 数据 ============================
norm_mean = [0.485, 0.456, 0.406]
norm_std = [0.229, 0.224, 0.225]
train_transform = transforms.Compose([
transforms.Resize(32),
transforms.RandomHorizontalFlip(p=0.5),
transforms.RandomCrop(32, padding=4),
transforms.ToTensor(),
transforms.Normalize(norm_mean, norm_std),
])
valid_transform = transforms.Compose([
transforms.Resize((32, 32)),
transforms.ToTensor(),
transforms.Normalize(norm_mean, norm_std),
])
# 构建MyDataset实例
train_data = CifarDataset(data_dir=train_dir, transform=train_transform)
valid_data = CifarDataset(data_dir=test_dir, transform=valid_transform)
# 构建DataLoder
train_loader = DataLoader(dataset=train_data, batch_size=BATCH_SIZE, shuffle=True, num_workers=2)
valid_loader = DataLoader(dataset=valid_data, batch_size=BATCH_SIZE, num_workers=2)
# ============================ step 2/5 模型 ============================
resnet_model = resnet20()
resnet_model.to(device)
# ============================ step 3/5 损失函数 ============================
criterion = nn.CrossEntropyLoss()
# ============================ step 4/5 优化器 ============================
# 冻结卷积层
optimizer = optim.SGD(resnet_model.parameters(), lr=LR, momentum=0.9, weight_decay=1e-4) # 选择优化器
scheduler = optim.lr_scheduler.MultiStepLR(optimizer, gamma=0.1, milestones=milestones)
# ============================ step 5/5 训练 ============================
loss_rec = {"train": [], "valid": []}
acc_rec = {"train": [], "valid": []}
best_acc, best_epoch = 0, 0
for epoch in range(start_epoch + 1, MAX_EPOCH):
# 训练(data_loader, model, loss_f, optimizer, epoch_id, device, max_epoch)
loss_train, acc_train, mat_train = ModelTrainer.train(train_loader, resnet_model, criterion, optimizer, epoch, device, MAX_EPOCH)
loss_valid, acc_valid, mat_valid = ModelTrainer.valid(valid_loader, resnet_model, criterion, device)
print("Epoch[{:0>3}/{:0>3}] Train Acc: {:.2%} Valid Acc:{:.2%} Train loss:{:.4f} Valid loss:{:.4f} LR:{}".format(
epoch + 1, MAX_EPOCH, acc_train, acc_valid, loss_train, loss_valid, optimizer.param_groups[0]["lr"]))
scheduler.step() # 更新学习率
# 绘图
loss_rec["train"].append(loss_train), loss_rec["valid"].append(loss_valid)
acc_rec["train"].append(acc_train), acc_rec["valid"].append(acc_valid)
show_confMat(mat_train, class_names, "train", log_dir, verbose=epoch == MAX_EPOCH-1)
show_confMat(mat_valid, class_names, "valid", log_dir, verbose=epoch == MAX_EPOCH-1)
plt_x = np.arange(1, epoch+2)
plot_line(plt_x, loss_rec["train"], plt_x, loss_rec["valid"], mode="loss", out_dir=log_dir)
plot_line(plt_x, acc_rec["train"], plt_x, acc_rec["valid"], mode="acc", out_dir=log_dir)
if epoch > (MAX_EPOCH/2) and best_acc < acc_valid:
best_acc = acc_valid
best_epoch = epoch
checkpoint = {"model_state_dict": resnet_model.state_dict(),
"optimizer_state_dict": optimizer.state_dict(),
"epoch": epoch,
"best_acc": best_acc}
path_checkpoint = os.path.join(log_dir, "checkpoint_best.pkl")
torch.save(checkpoint, path_checkpoint)
print(" done ~~~~ {}, best acc: {} in :{} epochs. ".format(datetime.strftime(datetime.now(), '%m-%d_%H-%M'),
best_acc, best_epoch))
now_time = datetime.now()
time_str = datetime.strftime(now_time, '%m-%d_%H-%M')
print(time_str)
四、Resnet架构介绍
class ResNet(nn.Module):
def __init__(self, block, layers, num_classes=1000, zero_init_residual=False,
groups=1, width_per_group=64, replace_stride_with_dilation=None,
norm_layer=None):
super(ResNet, self).__init__()
# 设置标准化层,默认为BN
if norm_layer is None:
norm_layer = nn.BatchNorm2d
self._norm_layer = norm_layer
self.inplanes = 64 # 第一个卷积卷积核数量,后续各个stage,用planes控制通道数,64-128-256-512
self.dilation = 1
if replace_stride_with_dilation is None:
# each element in the tuple indicates if we should replace
# the 2x2 stride with a dilated convolution instead
replace_stride_with_dilation = [False, False, False]
if len(replace_stride_with_dilation) != 3:
raise ValueError("replace_stride_with_dilation should be None "
"or a 3-element tuple, got {}".format(replace_stride_with_dilation))
self.groups = groups
self.base_width = width_per_group # 64
# stage: 1
self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3,
bias=False)
self.bn1 = norm_layer(self.inplanes)
self.relu = nn.ReLU(inplace=True)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
# stage: 2-5
self.layer1 = self._make_layer(block, 64, layers[0]) # basic, 64, 2
self.layer2 = self._make_layer(block, 128, layers[1], stride=2,
dilate=replace_stride_with_dilation[0]) # basic, 128, 2, 2, false
self.layer3 = self._make_layer(block, 256, layers[2], stride=2,
dilate=replace_stride_with_dilation[1]) # basic, 256, 2, 2, false
self.layer4 = self._make_layer(block, 512, layers[3], stride=2,
dilate=replace_stride_with_dilation[2]) # basic, 512, 2, 2, false
# stage:6
self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
self.fc = nn.Linear(512 * block.expansion, num_classes) # block.expansion=1
# 初始化
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.GroupNorm)):
nn.init.constant_(m.weight, 1)
nn.init.constant_(m.bias, 0)
# Zero-initialize the last BN in each residual branch,
# so that the residual branch starts with zeros, and each residual block behaves like an identity.
# This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677
if zero_init_residual:
for m in self.modules():
if isinstance(m, Bottleneck):
nn.init.constant_(m.bn3.weight, 0)
elif isinstance(m, BasicBlock):
nn.init.constant_(m.bn2.weight, 0)
def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
norm_layer = self._norm_layer # nn.BatchNorm2d
downsample = None
previous_dilation = self.dilation
if dilate:
self.dilation *= stride
stride = 1
# shortcut connection 的B策略,当分辨率变化时,采用1*1卷积进行变换特征图分辨率
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
conv1x1(self.inplanes, planes * block.expansion, stride),
norm_layer(planes * block.expansion),
)
layers = []
# 添加第一个building block,由于特征图分辨率下降在第一个building block中进行,因此这一个block比较特别,单独拿出来添加
layers.append(block(self.inplanes, planes, stride, downsample, self.groups,
self.base_width, previous_dilation, norm_layer))
# 更新 self.inplanes
self.inplanes = planes * block.expansion
# 添加其余building block
for _ in range(1, blocks):
layers.append(block(self.inplanes, planes, groups=self.groups,
base_width=self.base_width, dilation=self.dilation,
norm_layer=norm_layer))
return nn.Sequential(*layers)
def _forward_impl(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 = torch.flatten(x, 1)
x = self.fc(x)
return x
def forward(self, x):
return self._forward_impl(x)
#%%
def _resnet(arch, block, layers, pretrained, progress, **kwargs):
model = ResNet(block, layers, **kwargs)
if pretrained:
state_dict = load_state_dict_from_url(model_urls[arch],
progress=progress)
model.load_state_dict(state_dict)
return model
两种残差结构定义
def conv1x1(in_planes, out_planes, stride=1):
"""1x1 convolution"""
return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)
class BasicBlock(nn.Module):
expansion = 1 # 最后FC层使用
__constants__ = ['downsample']
def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
base_width=64, dilation=1, norm_layer=None):
super(BasicBlock, self).__init__()
if norm_layer is None:
norm_layer = nn.BatchNorm2d
if groups != 1 or base_width != 64:
raise ValueError('BasicBlock only supports groups=1 and base_width=64')
if dilation > 1:
raise NotImplementedError("Dilation > 1 not supported in BasicBlock")
# Both self.conv1 and self.downsample layers downsample the input when stride != 1
self.conv1 = conv3x3(inplanes, planes, stride)
self.bn1 = norm_layer(planes)
self.relu = nn.ReLU(inplace=True)
self.conv2 = conv3x3(planes, planes)
self.bn2 = norm_layer(planes)
self.downsample = downsample
self.stride = stride
def forward(self, x):
identity = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
if self.downsample is not None:
identity = self.downsample(x)
out += identity
out = self.relu(out)
return out
class Bottleneck(nn.Module):
expansion = 4
__constants__ = ['downsample']
def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
base_width=64, dilation=1, norm_layer=None):
super(Bottleneck, self).__init__()
if norm_layer is None:
norm_layer = nn.BatchNorm2d
width = int(planes * (base_width / 64.)) * groups
# Both self.conv2 and self.downsample layers downsample the input when stride != 1
self.conv1 = conv1x1(inplanes, width)
self.bn1 = norm_layer(width)
self.conv2 = conv3x3(width, width, stride, groups, dilation)
self.bn2 = norm_layer(width)
self.conv3 = conv1x1(width, planes * self.expansion)
self.bn3 = norm_layer(planes * self.expansion)
self.relu = nn.ReLU(inplace=True)
self.downsample = downsample
self.stride = stride
def forward(self, x):
identity = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
out = self.relu(out)
out = self.conv3(out)
out = self.bn3(out)
if self.downsample is not None:
identity = self.downsample(x)
out += identity
out = self.relu(out)
return out