PaddlePaddle中的圖像分類任務(wù)如何實(shí)現(xiàn)

小樊
69
2024-03-30 20:03:07

在PaddlePaddle中實(shí)現(xiàn)圖像分類任務(wù)通常使用卷積神經(jīng)網(wǎng)絡(luò)（CNN）。以下是一個(gè)簡(jiǎn)單的圖像分類示例：

  1. 導(dǎo)入必要的庫(kù)和模塊：
import paddle
import paddle.nn.functional as F
from paddle.vision import transforms
  1. 定義一個(gè)簡(jiǎn)單的卷積神經(jīng)網(wǎng)絡(luò)模型：
class Net(paddle.nn.Layer):
    def __init__(self, num_classes=10):
        super(Net, self).__init__()
        self.conv1 = paddle.nn.Conv2D(in_channels=3, out_channels=32, kernel_size=3, stride=1, padding=1)
        self.pool1 = paddle.nn.MaxPool2D(kernel_size=2, stride=2)
        self.conv2 = paddle.nn.Conv2D(in_channels=32, out_channels=64, kernel_size=3, stride=1, padding=1)
        self.pool2 = paddle.nn.MaxPool2D(kernel_size=2, stride=2)
        self.fc1 = paddle.nn.Linear(in_features=64*8*8, out_features=128)
        self.fc2 = paddle.nn.Linear(in_features=128, out_features=num_classes)

    def forward(self, x):
        x = self.pool1(F.relu(self.conv1(x)))
        x = self.pool2(F.relu(self.conv2(x)))
        x = paddle.flatten(x, start_axis=1)
        x = F.relu(self.fc1(x))
        x = self.fc2(x)
        return x
  1. 準(zhǔn)備數(shù)據(jù)和數(shù)據(jù)增強(qiáng)：
transform = transforms.Compose([
    transforms.Resize(size=32),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor()
])

train_dataset = paddle.vision.datasets.CIFAR10(mode='train', transform=transform)
train_loader = paddle.io.DataLoader(train_dataset, batch_size=32, shuffle=True)

test_dataset = paddle.vision.datasets.CIFAR10(mode='test', transform=transform)
test_loader = paddle.io.DataLoader(test_dataset, batch_size=32, shuffle=False)
  1. 訓(xùn)練模型：
model = Net()
optimizer = paddle.optimizer.Adam(parameters=model.parameters())
criterion = paddle.nn.CrossEntropyLoss()

model.train()

for epoch in range(10):
    for data in train_loader:
        images, labels = data
        outputs = model(images)
        loss = criterion(outputs, labels)
        
        optimizer.clear_grad()
        loss.backward()
        optimizer.step()
  1. 在測(cè)試集上評(píng)估模型：
model.eval()

accs = []
for data in test_loader:
    images, labels = data
    outputs = model(images)
    acc = paddle.metric.accuracy(outputs, labels)
    accs.append(acc.numpy())
    
print("Test Accuracy: ", sum(accs) / len(accs))

這是一個(gè)簡(jiǎn)單的圖像分類示例，實(shí)際應(yīng)用中可以根據(jù)需求調(diào)整網(wǎng)絡(luò)結(jié)構(gòu)、數(shù)據(jù)增強(qiáng)方式、優(yōu)化器等參數(shù)進(jìn)行優(yōu)化。

0 踩