AlexNet复现代码

train.py

import torch
import torch.nn as nn
import torch.nn.functional as F
import os
from tensorboardX import SummaryWriter
import torchvision.datasets as Datasets
import torchvision.transforms as transforms
import torch.utils.data.dataloader as Dataloader
from tqdm import tqdm

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

NUM_EPOCHS = 10
BATCH_SIZE = 4
MOMENTUM = 0.9
LR_DECAY = 0.0005
LR_INIT = 0.01
IMAGE_DIM = 227
NUM_CLASSES = 10
DEVICE_DIS = [0, 1, 2, 3]



#数据集和输出路径
INPUT_ROOT_DIR = 'alexnet_data_in'
TRAIN_IMG_DIR = INPUT_ROOT_DIR + '/imagenet'
OUTPUT_DIR = 'alexnet_data_out'
LOG_DIR = OUTPUT_DIR + '/tblogs'
CHECKPOINT_DIR = OUTPUT_DIR + '/models'

os.makedirs(CHECKPOINT_DIR, exist_ok = True)



class AlexNet(nn.Module):
    def __init__(self):
        super(AlexNet, self).__init__()
        self.conv1 = nn.Conv2d(3, 96, 11, stride = 4)
        self.conv2 = nn.Conv2d(96, 256, 5, padding = 2)
        self.conv3 = nn.Conv2d(256, 384, 2, padding = 1)
        self.conv4 = nn.Conv2d(384, 384, 3, padding = 1)
        self.conv5 = nn.Conv2d(384, 256, 3, padding = 1)
        self.fc1 = nn.Linear((256 * 6 * 6), 4096)
        self.fc2 = nn.Linear(4096, 4096)
        self.fc3 = nn.Linear(4096, NUM_CLASSES)
    def forward(self, x):
        x = F.max_pool2d(F.local_response_norm(F.relu(self.conv1(x)), size = 5, alpha = 0.0001, beta = 0.75, k = 2), kernel_size = 3, stride = 2)
        x = F.max_pool2d(F.local_response_norm(F.relu(self.conv2(x)), size = 5, alpha = 0.0001, beta = 0.75, k = 2), kernel_size = 3, stride = 2)
        x = F.relu(self.conv3(x))
        x = F.relu(self.conv4(x))
        x = F.max_pool2d(F.relu(self.conv5(x)), kernel_size = 3, stride = 2)
        x = x.view(x.size()[0], -1)
        x = F.relu(self.fc1(x))
        x = F.dropout(x, p = 0.5, training = True, inplace = True)
        x = F.relu(self.fc2(x))
        x = F.dropout(x, p = 0.5, training = True, inplace = True)
        x = self.fc3(x)
        return x



if __name__ == '__main__':

    tbwrite = SummaryWriter(log_dir = LOG_DIR)
    print('tensorboardX summary write created')
    alexnet = AlexNet().to(device)
    print(alexnet)
    print('alexnet created')
    dataset = Datasets.ImageFolder(TRAIN_IMG_DIR, transform = transforms.Compose([
        transforms.CenterCrop(IMAGE_DIM),
        #将读进来的图片转换为tensor
        transforms.ToTensor(),
        #对tensor进行归一化
        transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
    ]))
    print('dataset created')

    dataloader = Dataloader.DataLoader(
        dataset,
        shuffle = True,
        num_workers = 8,
        drop_last = True,
        batch_size = BATCH_SIZE)
    print('dataloader created')
    optimizer = torch.optim.Adam(alexnet.parameters(), lr = 0.0001)
    #每30轮将lr * 0.1
    lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size = 30, gamma = 0.1)
    loss = nn.CrossEntropyLoss()
    print('Start Training...')
    cnt = 0
    for epoch in range(NUM_EPOCHS):
        true_total = 0
        img_total = 0
        loss_total = 0
        for imgs, classes in tqdm(dataloader):
            cnt += 1
            optimizer.zero_grad()
            imgs, classes = imgs.to(device), classes.to(device)
            print(classes)
            output = alexnet(imgs)
            loss_value = loss(output, classes)
            loss_value.backward()
            optimizer.step()

            preds = torch.max(output, 1)[1]
            true_total += torch.sum(preds == classes)
            img_total += len(classes)
            loss_total += loss_value.item()

        loss_k = float(loss_total) / float(cnt)
        accuracy = float(true_total) / float(img_total)
        print('epoch: {} 	 Loss: {:.4f} 	 Acc: {:.2f}'.format(epoch + 1, loss_k, accuracy))
        tbwrite.add_scalar('loss', loss_k, epoch + 1)
        tbwrite.add_scalar('accuracy', accuracy, epoch + 1)



        #log information and add to tensorboard

        lr_scheduler.step()


        if (epoch + 1) % 10 == 0:
            checkpoint_path = os.path.join(CHECKPOINT_DIR, 'alexnet_staes_e{}.pkl'.format(epoch + 1))
            state = {
                'epoch' : epoch,
                'optimizer' : optimizer.state_dict(),
                'model' : alexnet.state_dict(),
            }
            torch.save(state, checkpoint_path)

predict.py

import torch
import torch.nn
import torch.nn.functional as F
import numpy as np
import torchvision.datasets as Dataset
import torchvision.transforms as transforms
import torch.utils.data.dataloader as Dataloader
from PIL import Image
import matplotlib.pyplot as plt
import os
from train import AlexNet
import cv2

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
IMG_DIM = 227
INPUT_ROOT_DIR = 'alexnet_data_in'
OUTPUT_ROOT_DIR = 'alexnet_data_out'
IMG_DIR = INPUT_ROOT_DIR + '/test_imagenet'

checkpoint_dir = os.path.join(OUTPUT_ROOT_DIR, 'models', 'alexnet_staes_e10.pkl')
checkpoint = torch.load(checkpoint_dir)
model = AlexNet()
model.to(device)
model.load_state_dict(checkpoint['model'])


def open_list(dir):
    for home, files, dirs in os.walk(dir):
        return dirs

def show(name, img):
    cv2.imshow(name, img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()


transform = transforms.Compose([
    transforms.CenterCrop(IMG_DIM),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])]
)


img_list = open_list(IMG_DIR)


for img_name in img_list:
    img_path = os.path.join(IMG_DIR, img_name)
    img = Image.open(img_path)
    img = transform(img)
    img.to(device)
    #注意要reshape一下，module是默认有批次的
    img = img.reshape([-1, 3, 227, 227])
    preds = model(img)
    preds = torch.max(preds, 1)[1]
    print(preds.item())
    img = cv2.imread(img_path)
    show('{}'.format(preds.item()), img)