1.Lenet网络简介
LeNet是一种经典的卷积神经网络(Convolutional Neural Network,CNN),也被称为LeNet-5。它由Yann LeCun等人在1998年提出,是用于手写数字识别任务的早期深度学习模型。
LeNet网络结构相对简单,主要由卷积层、池化层和全连接层组成。它的主要特点是通过多层卷积和池化操作对输入进行特征提取,并使用全连接层进行分类。具体来说,LeNet的结构如下:
- 输入层:接受图像输入。
- 卷积层:使用卷积核对输入进行卷积操作,并应用非线性激活函数(如sigmoid或tanh)。
- 池化层:对卷积输出进行下采样操作,通常使用平均池化或最大池化。
- 卷积层和池化层:可以重复几次,以提取更高级别的特征。
- 全连接层:将池化层的输出连接到一个或多个全连接层,最后输出预测结果。
- 输出层:根据任务需求确定输出层的激活函数,如softmax用于多类别分类问题。
LeNet在当时取得了很好的效果,奠定了卷积神经网络在计算机视觉领域的基础。它的设计和思想对之后更复杂的深度学习模型的发展有着重要影响。
值得一提的是,虽然LeNet网络在手写数字识别任务上表现出色,但在处理更复杂的图像任务(如物体识别)时可能效果较差。随着技术的发展,更深层次的卷积神经网络(如VGG、ResNet、Inception等)被提出并取得了显著的进展。
2.Lenet网络实现
代码是参考某站博主“跟李沐学ai”,重写了d2l中的train_ch6方法,将测试集和训练集精度直接输出。
import torch
from d2l.torch import evaluate_accuracy_gpu
from torch import nn
from d2l import torch as d2l
import cv2
def train_ch6(net, train_iter, test_iter, num_epochs, lr, device):
def init_weights(m):
if type(m) == nn.Linear or type(m) == nn.Conv2d:
nn.init.xavier_uniform_(m.weight)
net.apply(init_weights)
print('training on', device)
net.to(device)
optimizer = torch.optim.SGD(net.parameters(), lr=lr)
loss = nn.CrossEntropyLoss()
timer, num_batches = d2l.Timer(), len(train_iter)
for epoch in range(num_epochs):
# Sum of training loss, sum of training accuracy, no. of examples
metric = d2l.Accumulator(3)
net.train()
for i, (X, y) in enumerate(train_iter):
timer.start()
optimizer.zero_grad()
X, y = X.to(device), y.to(device)
y_hat = net(X)
l = loss(y_hat, y)
l.backward()
optimizer.step()
with torch.no_grad():
metric.add(l * X.shape[0], d2l.accuracy(y_hat, y), X.shape[0])
timer.stop()
train_l = metric[0] / metric[2]
train_acc = metric[1] / metric[2]
test_acc = evaluate_accuracy_gpu(net, test_iter)
print(f'epoch {epoch},loss :{train_l:.3f}, train acc :{train_acc:.3f}, '
f'test acc :{test_acc:.3f}, time :{timer.sum() / 60}m{int(timer.sum() % 60)}s')
print(f'loss {train_l:.3f}, train acc {train_acc:.3f}, '
f'test acc {test_acc:.3f}')
print(f'{metric[2] * num_epochs / timer.sum():.1f} examples/sec '
f'on {str(device)}')
if __name__ == "__main__":
# 模型训练
net = nn.Sequential(nn.Conv2d(1, 6, kernel_size=5, padding=2), nn.Sigmoid(),
nn.AvgPool2d(kernel_size=2, stride=2),
nn.Conv2d(6, 16, kernel_size=5), nn.Sigmoid(),
nn.AvgPool2d(kernel_size=2, stride=2), nn.Flatten(),
nn.Linear(16 * 5 * 5, 120), nn.Sigmoid(),
nn.Linear(120, 84), nn.Sigmoid(),
nn.Linear(84, 10))
# net.load_state_dict(torch.load('lenet/lenet15.pth'))
batch_size = 256
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size=batch_size)
lr, num_epochs = 0.9, 50
train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())
# 模型保存
torch.save(net.state_dict(), f'Lenet/Lenet{num_epochs}.pth')
# 加载模型
# net = nn.Sequential(nn.Conv2d(1, 6, kernel_size=5, padding=2), nn.Sigmoid(),
# nn.AvgPool2d(kernel_size=2, stride=2),
# nn.Conv2d(6, 16, kernel_size=5), nn.Sigmoid(),
# nn.AvgPool2d(kernel_size=2, stride=2), nn.Flatten(),
# nn.Linear(16 * 5 * 5, 120), nn.Sigmoid(),
# nn.Linear(120, 84), nn.Sigmoid(),
# nn.Linear(84, 10))
# net.load_state_dict(torch.load('lenet.pth'))
# 预测图片预处理
def preprocess(image_path):
image = cv2.cvtColor(cv2.imread(image_path), cv2.COLOR_BGR2GRAY)
image = torch.tensor(cv2.resize(image, (28, 28)), dtype=torch.float32).unsqueeze(0).unsqueeze(0)
return image
# 预测图片
def predict(image_path):
image = preprocess(image_path)
# TODO 若是训练完直接预测,由于网络是在GPU上,因此需要将输入张量转到GPU上
image = image.cuda()
with torch.no_grad():
outputs = net(image)
_, predicted = torch.max(outputs)
# 这里的class_label需要自己修改
class_label = d2l.get_fashion_mnist_labels(predicted)
return class_label
image_path = 'Lenet/777.png' # 待预测的图像路径
predicted_class = predict(image_path)
print('预测类别:', predicted_class)