寫在前面： 我是「虐貓人薛定諤i」，一個不滿足於現狀，有夢想，有追求的00後
本博客主要記錄和分享自己畢生所學的知識，歡迎關注，第一時間獲取更新。
不忘初心，方得始終。
❤❤❤❤❤❤❤❤❤❤
LeNet簡介
LeNet是最早的卷積神經網絡之一，其網絡結構如下圖所示
從圖中我們可以看到，該網絡包含3個卷積層，2個池化層和2個全連接層，LeNet通過連續使用卷積層和全連接層來提取圖像的特徵，進而達到識別圖像的目的。
代碼
LeNet網絡的實現代碼如下
class LeNet(fluid.dygraph.Layer):
def __init__(self, name_scope, num_classes=1):
super(LeNet, self).__init__(name_scope)
self.conv1 = Conv2D(num_channels=1,
num_filters=6,
filter_size=5,
act='sigmoid')
self.pool1 = Pool2D(pool_size=2, pool_stride=2, pool_type='max')
self.conv2 = Conv2D(num_channels=6,
num_filters=16,
filter_size=5,
act='sigmoid')
self.pool2 = Pool2D(pool_size=2, pool_stride=2, pool_type='max')
self.conv3 = Conv2D(num_channels=16,
num_filters=120,
filter_size=4,
act='sigmoid')
self.fc1 = Linear(input_dim=120, output_dim=64, act='sigmoid')
self.fc2 = Linear(input_dim=64, output_dim=num_classes)
def forward(self, x):
x = self.conv1(x)
x = self.pool1(x)
x = self.conv2(x)
x = self.pool2(x)
x = self.conv3(x)
x = fluid.layers.reshape(x, [x.shape[0], -1])
x = self.fc1(x)
x = self.fc2(x)
return x
完整代碼如下
import paddle
import paddle.fluid as fluid
import numpy as np
from paddle.fluid.dygraph.nn import Conv2D, Pool2D, Linear
"""
LeNet在手寫數字識別上的應用
"""
class LeNet(fluid.dygraph.Layer):
def __init__(self, name_scope, num_classes=1):
super(LeNet, self).__init__(name_scope)
self.conv1 = Conv2D(num_channels=1,
num_filters=6,
filter_size=5,
act='sigmoid')
self.pool1 = Pool2D(pool_size=2, pool_stride=2, pool_type='max')
self.conv2 = Conv2D(num_channels=6,
num_filters=16,
filter_size=5,
act='sigmoid')
self.pool2 = Pool2D(pool_size=2, pool_stride=2, pool_type='max')
self.conv3 = Conv2D(num_channels=16,
num_filters=120,
filter_size=4,
act='sigmoid')
self.fc1 = Linear(input_dim=120, output_dim=64, act='sigmoid')
self.fc2 = Linear(input_dim=64, output_dim=num_classes)
def forward(self, x):
x = self.conv1(x)
x = self.pool1(x)
x = self.conv2(x)
x = self.pool2(x)
x = self.conv3(x)
x = fluid.layers.reshape(x, [x.shape[0], -1])
x = self.fc1(x)
x = self.fc2(x)
return x
def train(model):
print("start training...")
model.train()
epoch_num = 5
opt = fluid.optimizer.Momentum(learning_rate=0.001,
momentum=0.9,
parameter_list=model.parameters())
train_loader = paddle.batch(paddle.dataset.mnist.train(), batch_size=10)
valid_loader = paddle.batch(paddle.dataset.mnist.test(), batch_size=10)
for epoch in range(epoch_num):
for batch_id, data in enumerate(train_loader()):
x_data = np.array([item[0] for item in data],
dtype='float32').reshape(-1, 1, 28, 28)
y_data = np.array([item[1] for item in data],
dtype='int64').reshape(-1, 1)
img = fluid.dygraph.to_variable(x_data)
label = fluid.dygraph.to_variable(y_data)
logits = model(img)
loss = fluid.layers.softmax_with_cross_entropy(logits, label)
avg_loss = fluid.layers.mean(loss)
if batch_id % 1000 == 0:
print("epoch: {}, bath_id: {}, loss is: {}".format(
epoch, batch_id, avg_loss.numpy()))
avg_loss.backward()
opt.minimize(avg_loss)
model.clear_gradients()
model.eval()
accuracies = []
losses = []
for batch_id, data in enumerate(valid_loader()):
x_data = np.array([item[0] for item in data],
dtype='float32').reshape(-1, 1, 28, 28)
y_data = np.array([item[1] for item in data],
dtype='int64').reshape(-1, 1)
img = fluid.dygraph.to_variable(x_data)
label = fluid.dygraph.to_variable(y_data)
logits = model(img)
pred = fluid.layers.softmax(logits)
loss = fluid.layers.softmax_with_cross_entropy(logits, label)
acc = fluid.layers.accuracy(pred, label)
accuracies.append(acc.numpy())
losses.append(loss.numpy())
print("[validation accuracy/loss: {}/{}]".format(
np.mean(accuracies), np.mean(losses)))
model.train()
fluid.save_dygraph(model.state_dict(), './result/hwdrByLeNet')
if __name__ == '__main__':
with fluid.dygraph.guard():
model = LeNet('LeNet', num_classes=10)
train(model)
結果
start training...
epoch: 0, bath_id: 0, loss is: [2.2495162]
epoch: 0, bath_id: 1000, loss is: [2.2928371]
epoch: 0, bath_id: 2000, loss is: [2.3267434]
epoch: 0, bath_id: 3000, loss is: [2.2698295]
epoch: 0, bath_id: 4000, loss is: [2.2489858]
epoch: 0, bath_id: 5000, loss is: [2.312758]
[validation accuracy/loss: 0.45590001344680786/2.215536117553711]
epoch: 1, bath_id: 0, loss is: [2.1956322]
epoch: 1, bath_id: 1000, loss is: [2.063491]
epoch: 1, bath_id: 2000, loss is: [1.9574039]
epoch: 1, bath_id: 3000, loss is: [1.420162]
epoch: 1, bath_id: 4000, loss is: [0.98229045]
epoch: 1, bath_id: 5000, loss is: [1.2404814]
[validation accuracy/loss: 0.776199996471405/0.8473402261734009]
epoch: 2, bath_id: 0, loss is: [0.62948656]
epoch: 2, bath_id: 1000, loss is: [0.49548474]
epoch: 2, bath_id: 2000, loss is: [0.5145985]
epoch: 2, bath_id: 3000, loss is: [0.2760195]
epoch: 2, bath_id: 4000, loss is: [0.36493483]
epoch: 2, bath_id: 5000, loss is: [0.5631878]
[validation accuracy/loss: 0.8793999552726746/0.4475659728050232]
epoch: 3, bath_id: 0, loss is: [0.30772734]
epoch: 3, bath_id: 1000, loss is: [0.2511763]
epoch: 3, bath_id: 2000, loss is: [0.32035473]
epoch: 3, bath_id: 3000, loss is: [0.12164386]
epoch: 3, bath_id: 4000, loss is: [0.20446599]
epoch: 3, bath_id: 5000, loss is: [0.27960077]
[validation accuracy/loss: 0.9111999869346619/0.3133259415626526]
epoch: 4, bath_id: 0, loss is: [0.16361086]
epoch: 4, bath_id: 1000, loss is: [0.15575354]
epoch: 4, bath_id: 2000, loss is: [0.24734934]
epoch: 4, bath_id: 3000, loss is: [0.07145926]
epoch: 4, bath_id: 4000, loss is: [0.14044744]
epoch: 4, bath_id: 5000, loss is: [0.16796467]
[validation accuracy/loss: 0.9281999468803406/0.24646225571632385]
總結
通過最後的結果，我們可以看到手寫數字識別的準去率達到了92.8%
蒟蒻寫博客不易，加之本人水平有限，寫作倉促，錯誤和不足之處在所難免，謹請讀者和各位大佬們批評指正。
如需轉載，請署名作者並附上原文鏈接，蒟蒻非常感激
名稱：虐貓人薛定諤i
博客地址：https://blog.csdn.net/Deep___Learning