📝Docs: [Lab-5] Completed

.gitignore

@@ -1,2 +1,8 @@
 .idea
-*/img
+*/img
+*/data
+model
+runs
+datasets
+yolov8n.pt
+dataset

Lab-2/src/data_load.py

@@ -0,0 +1,67 @@
+import torch
+import torchvision
+import matplotlib.pyplot as plt
+from torchvision import transforms
+from torch.utils.data import random_split, DataLoader
+
+
+def load_and_preview_mnist(data_size=100, batch_size=16, seed=42, root='../data'):
+    """
+    加载并预览 MNIST 数据集。
+    Parameters:
+    - data_size (int): 加载的数据条目数。
+    - batch_size (int): 每批次处理的数据数量。
+    - seed (int): 随机种子，用于保证数据切割的可重复性。
+    - root (str): 数据集保存的路径。
+    """
+    # 数据预处理
+    transform = transforms.Compose([
+        transforms.ToTensor(),
+        transforms.Normalize((0.5,), (0.5,))
+    ])
+
+    # 载入数据集
+    train_dataset = torchvision.datasets.MNIST(
+        root=root,
+        train=True,
+        download=True,
+        transform=transform
+    )
+
+    # 截取指定数量的数据
+    data_train, _ = random_split(
+        dataset=train_dataset,
+        lengths=[data_size, len(train_dataset) - data_size],
+        generator=torch.Generator().manual_seed(seed)
+    )
+
+    # 加载数据集
+    train_loader = DataLoader(
+        dataset=data_train,
+        batch_size=batch_size,
+        shuffle=True
+    )
+
+    # 数据预览
+    images, labels = next(iter(train_loader))
+
+    # 将图像拼接为网格
+    img_grid = torchvision.utils.make_grid(images, nrow=batch_size // 4)
+    img = img_grid.numpy().transpose(1, 2, 0)
+    std = [0.5]
+    mean = [0.5]
+    img = img * std + mean
+
+    # 显示图像
+    print([labels[i] for i in range(batch_size)])
+    plt.imshow(img, cmap='gray')
+    plt.axis('off')
+    plt.show()
+
+
+if __name__ == '__main__':
+    # 使用默认参数调用函数
+    load_and_preview_mnist()
+
+    # 或者使用自定义参数
+    # load_and_preview_mnist(data_size=200, batch_size=8, seed=123, root='./data')

Lab-2/src/data_transform.py

@@ -0,0 +1,89 @@
+import torch
+import torchvision
+import torchvision.transforms as transforms
+import matplotlib.pyplot as plt
+from torch.utils.data import DataLoader, random_split
+
+
+def get_transform():
+    """
+    返回 torchvision.transforms.Compose 对象，包含数据预处理变换操作。
+
+    Returns:
+    - transform (torchvision.transforms.Compose): 包含数据预处理变换的 Compose 对象。
+    """
+    return transforms.Compose([
+        transforms.ToTensor(),  # 将输入数据类型转换为 Tensor
+        transforms.Normalize((0.5,), (0.5,))  # 对数据进行标准化处理，均值和标准差均为 0.5
+    ])
+
+
+def load_mnist(data_size=100, batch_size=16, seed=42, root='../data', train=True):
+    """
+    加载 MNIST 数据集。
+
+    Parameters:
+    - data_size (int): 加载的数据条目数。
+    - batch_size (int): 每批次处理的数据数量。
+    - seed (int): 随机种子，用于保证数据切割的可重复性。
+    - root (str): 数据集保存的路径。
+    - train (bool): 是否加载训练集。
+
+    Returns:
+    - DataLoader: 数据加载器。
+    """
+    transform = get_transform()
+
+    dataset = torchvision.datasets.MNIST(
+        root=root,
+        train=train,
+        download=True,
+        transform=transform
+    )
+
+    data_subset, _ = random_split(
+        dataset=dataset,
+        lengths=[data_size, len(dataset) - data_size],
+        generator=torch.Generator().manual_seed(seed)
+    )
+
+    data_loader = DataLoader(
+        dataset=data_subset,
+        batch_size=batch_size,
+        shuffle=True
+    )
+
+    return data_loader
+
+
+def load_and_preview_mnist(data_size=100, batch_size=16, seed=42, root='../data'):
+    """
+    加载并预览 MNIST 数据集。
+
+    Parameters:
+    - data_size (int): 加载的数据条目数。
+    - batch_size (int): 每批次处理的数据数量。
+    - seed (int): 随机种子，用于保证数据切割的可重复性。
+    - root (str): 数据集保存的路径。
+    """
+    train_loader = load_mnist(data_size=data_size, batch_size=batch_size, seed=seed, root=root, train=True)
+
+    images, labels = next(iter(train_loader))
+
+    img_grid = torchvision.utils.make_grid(images, nrow=batch_size // 4)
+    img = img_grid.numpy().transpose(1, 2, 0)
+    mean = [0.5]
+    std = [0.5]
+    img = img * std + mean
+
+    print([labels[i] for i in range(batch_size)])
+    plt.imshow(img, cmap='gray')
+    plt.axis('off')
+    plt.show()
+
+
+if __name__ == '__main__':
+    load_and_preview_mnist()
+
+    # Or use custom parameters:
+    # load_and_preview_mnist(data_size=200, batch_size=8, seed=123, root='./data')

Lab-2/src/model_training.py

@@ -0,0 +1,116 @@
+import torch
+import torch.nn as nn
+import torch.optim as optim
+from torch.utils.data import DataLoader
+from data_transform import load_mnist, load_and_preview_mnist
+import matplotlib.pyplot as plt
+import torchvision.utils as vutils
+
+
+# 定义卷积神经网络模型
+class Model(nn.Module):
+    def __init__(self):
+        super(Model, self).__init__()
+        self.conv1 = nn.Sequential(
+            nn.Conv2d(1, 64, kernel_size=3, stride=2, padding=1),
+            nn.ReLU(),
+            nn.Conv2d(64, 128, kernel_size=3, stride=2, padding=1),
+            nn.ReLU(),
+            # nn.MaxPool2d(stride=2, kernel_size=2)
+        )
+
+        self.dense = nn.Sequential(
+            nn.Linear(7 * 7 * 128, 512),
+            nn.ReLU(),
+            nn.Dropout(p=0.8),
+            nn.Linear(512, 10)
+        )
+
+    def forward(self, x):
+        x = self.conv1(x)  # 卷积处理
+        x = x.view(-1, 7 * 7 * 128)  # 对参数实行扁平化处理
+        x = self.dense(x)
+        return x
+
+
+# 加载和预览数据集
+def main():
+    # 参数设置
+    data_size = 1000
+    batch_size = 4
+    seed = 42
+    root = '../data'
+    epochs_n = 5
+    learning_rate = 0.001
+
+    # 加载数据加载器
+    train_loader = load_mnist(data_size=data_size, batch_size=batch_size, seed=seed, root=root, train=True)
+    test_loader = load_mnist(data_size=data_size, batch_size=batch_size, seed=seed, root=root, train=False)
+
+    # 预览数据集（可选）
+    load_and_preview_mnist(data_size=data_size, batch_size=batch_size, seed=seed, root=root)
+
+    # 初始化模型
+    model = Model()
+    cost = nn.CrossEntropyLoss()
+    optimizer = optim.Adam(model.parameters(), lr=learning_rate)
+
+    # 训练模型
+    for epoch in range(epochs_n):
+        model.train()
+        running_loss = 0.0
+        running_correct = 0
+        print(f"Epoch {epoch + 1}/{epochs_n}")
+        print("-" * 10)
+
+        for data in train_loader:
+            images, labels = data
+            optimizer.zero_grad()
+            outputs = model(images)
+            _, preds = torch.max(outputs.data, 1)
+            loss = cost(outputs, labels)
+            loss.backward()
+            optimizer.step()
+            running_loss += loss.item()
+            running_correct += (preds == labels).sum().item()
+
+        epoch_loss = running_loss / len(train_loader.dataset)
+        epoch_acc = 100 * running_correct / len(train_loader.dataset)
+
+        print(f"Loss: {epoch_loss:.4f}, Train Accuracy: {epoch_acc:.4f}%")
+
+        # 测试模型
+        model.eval()
+        testing_correct = 0
+        with torch.no_grad():
+            for data in test_loader:
+                images, labels = data
+                outputs = model(images)
+                _, preds = torch.max(outputs.data, 1)
+                testing_correct += (preds == labels).sum().item()
+
+            test_acc = 100 * testing_correct / len(test_loader.dataset)
+            print(f"Test Accuracy: {test_acc:.4f}%")
+
+    # 测试数据可视化
+    X_test, y_test = next(iter(test_loader))
+    with torch.no_grad():
+        pred = model(X_test)
+        _, predicted_labels = torch.max(pred, 1)
+
+    print("Predicted Labels:", [i.item() for i in predicted_labels])
+    print("True Labels:", [i.item() for i in y_test])
+
+    # 显示图像
+    img_grid = vutils.make_grid(X_test, nrow=batch_size // 4)
+    img = img_grid.numpy().transpose(1, 2, 0)
+    mean = [0.5]
+    std = [0.5]
+    img = img * std + mean
+    plt.imshow(img, cmap='gray')
+    plt.axis('off')
+    plt.show()
+
+
+if __name__ == '__main__':
+    main()

Lab-2/test/verify_env.py

@@ -0,0 +1,44 @@
+import unittest
+
+import torch
+import torchvision
+from torchvision import transforms
+import matplotlib.pyplot as plt
+
+
+class MyTestCase(unittest.TestCase):
+    def test_verify_env(self):
+        self.assertIsNotNone(torch.__version__)
+        print("== Env test pass ==\n"
+              f"[torch] {torch.__version__}\n")
+    def test_using_dataset(selfs):
+        transform = transforms.Compose([
+            transforms.ToTensor(),
+            transforms.Normalize((0.5,), (0.5,))
+        ])
+        test_dataset = torchvision.datasets.MNIST(
+            root='../data',
+            train=False,
+            download=True,
+            transform=transform
+        )
+        test_loader = torch.utils.data.DataLoader(
+            test_dataset,
+            batch_size=64,
+            shuffle=False
+        )
+        images, labels = next(iter(test_loader))
+        fig, axes = plt.subplots(2, 5, figsize=(10, 4))
+        for i, ax in enumerate(axes.flat):
+            ax.imshow(images[i][0], cmap='gray')
+            ax.set_title(f'Label: {labels[i]}')
+            ax.axis('off')
+
+        plt.tight_layout
+        plt.show()
+
+        print(f"Successfully load MNIST dataset and show 10 images")
+        return images, labels
+
+if __name__ == '__main__':
+    unittest.main()

Lab-3/src/train.py

@@ -0,0 +1,158 @@
+import torch
+import torchvision
+from torchvision import datasets, models, transforms
+import os
+from torch.autograd import Variable
+import matplotlib.pyplot as plt
+import time
+
+model_path = '../model/model_name.pth'
+model_params_path = '../model/params_name.pth'
+
+transform = transforms.Compose(
+    [transforms.CenterCrop(224),
+    transforms.ToTensor(),
+    transforms.Normalize([0.5,0.5,0.5],[0.5,0.5,0.5])
+    ]
+)
+
+data_dir = "../data"  # 更改为您自己的数据集路径
+
+data_transform = {
+    x:transforms.Compose(
+        [
+            transforms.Resize(256),          # Resize图像到256x256
+            transforms.CenterCrop(224),        # 中心裁剪到224x224
+            transforms.ToTensor(),
+            transforms.Normalize(
+                mean=[0.485, 0.456, 0.406],    # ImageNet数据集的 normalize 设置
+                std=[0.229, 0.224, 0.225]
+            )
+        ]
+    )
+    for x in ["train","valid"]
+}
+
+image_datasets = {
+    x:datasets.ImageFolder(
+        root=os.path.join(data_dir,x),        # 将输入参数中的两个名字拼接成一个完整的文件路径
+        transform=data_transform[x]
+    )
+    for x in ["train","valid"]
+}
+
+dataloader = {
+    x:torch.utils.data.DataLoader(
+        dataset=image_datasets[x],
+        batch_size=16,
+        shuffle=True
+    )
+    for x in ["train","valid"]
+}
+
+X_example, y_example = next(iter(dataloader["train"]))
+example_classes = image_datasets["train"].classes    # ['cat', 'dog']
+index_classes = image_datasets["train"].class_to_idx #{'cat': 0, 'dog': 1}
+
+Use_gpu = torch.cuda.is_available()
+model = models.resnet50(pretrained=True)
+print(model)
+
+for param in model.parameters():
+    param.requires_grad = False
+
+# 重新设计分类器的结构
+model.fc = torch.nn.Linear(2048,2)
+print(model)
+
+if Use_gpu:
+    model = model.cuda()
+
+loss_f = torch.nn.CrossEntropyLoss()
+optimizer = torch.optim.Adam(model.fc.parameters(), lr=0.00001)
+
+has_been_trained = os.path.isfile(model_path)
+if has_been_trained:
+    epoch_n = 0
+    model.load_state_dict(torch.load(model_params_path))
+else:
+    epoch_n = 1
+
+time_open = time.time()
+for epoch in range(epoch_n, epoch_n + 5):  # 训练5个epoch
+    print("Epoch {}/{}".format(epoch, epoch_n + 4))
+    print("-" * 10)
+
+    for phase in ["train", "valid"]:
+        if phase == "train":
+            print("Training...")
+            model.train(True)  # 启用 BatchNormalization 和 Dropout
+        else:
+            print("Validing...")
+            model.train(False)  # 不启用 BatchNormalization 和 Dropout
+
+        running_loss = 0.0
+        running_corrects = 0
+
+        for batch, data in enumerate(dataloader[phase], 1):
+            X, y = data
+            if Use_gpu:
+                X, y = Variable(X.cuda()), Variable(y.cuda())
+            else:
+                X, y = Variable(X), Variable(y)
+
+            y_pred = model(X)
+            _, pred = torch.max(y_pred.data, 1)
+
+            optimizer.zero_grad()
+            loss = loss_f(y_pred, y)
+
+            if phase == "train":
+                loss.backward()
+                optimizer.step()
+
+            running_loss += loss.item()
+            running_corrects += torch.sum(pred == y.data)
+
+            if batch % 50 == 0 and phase == "train":
+                print("Batch {}, Train Loss:{:.4f},Train ACC:{:.4f}%".format(
+                    batch, running_loss / batch, 100.0 * running_corrects / (16 * batch)
+                )
+                )
+
+        epoch_loss = running_loss * 16 / len(image_datasets[phase])
+        epoch_acc = 100.0 * running_corrects / len(image_datasets[phase])
+
+        print("{} Loss:{:.4f} Acc:{:.4f}%".format(phase, epoch_loss, epoch_acc))
+
+        # 保存模型参数
+        best_acc = 0.0
+        if phase == 'valid' and epoch_acc > best_acc:
+            best_acc = epoch_acc
+            torch.save(model.state_dict(), model_params_path)
+
+time_end = time.time() - time_open
+print("程序运行时间:{}分钟...".format(int(time_end / 60)))
+
+# 读取训练集中的一个batch
+X_example, Y_example = next(iter(dataloader['train']))
+
+if Use_gpu:
+    X_example, Y_example = Variable(X_example.cuda()), Variable(Y_example.cuda())
+else:
+    X_example, Y_example = Variable(X_example), Variable(Y_example)
+
+y_pred = model(X_example)
+
+index_classes = image_datasets['train'].class_to_idx   # 显示类别对应的索引
+example_classes = image_datasets['train'].classes     # 将原始图像的类别保存起来
+
+img = torchvision.utils.make_grid(X_example)
+img = img.cpu().numpy().transpose([1,2,0])
+
+print("实际:",[example_classes[i] for i in Y_example])
+_, y_pred = torch.max(y_pred.data,1)
+print("预测:",[example_classes[i] for i in y_pred])
+
+plt.imshow(img)
+plt.show()

Lab-4/src/detect.py

@@ -0,0 +1,20 @@
+from ultralytics import YOLO
+import cv2
+
+# 加载最佳模型
+model = YOLO('runs/detect/yolov8n_coco128_exp1/weights/best.pt')
+
+# 在图像上进行推理
+source = 'datasets/img3.png'  # 替换成你自己的测试图片路径
+results = model(source, save=True, conf=0.5)  # conf为置信度阈值
+
+# 可视化结果（可选，使用OpenCV）
+# 读取带标注的结果图像
+result_image = results[0].plot()  # 获取第一张图片的绘制结果
+cv2.imshow('Detection Result', result_image)
+cv2.waitKey(0)
+cv2.destroyAllWindows()
+
+# 也可以在视频或摄像头上进行推理
+# results = model('path/to/video.mp4', save=True, show=True)
+# results = model(0, show=True)  # 0代表默认摄像头

Lab-4/src/train.py

@@ -0,0 +1,15 @@
+from ultralytics import YOLO
+
+# 1. 加载一个预训练模型（小规模模型，适合实验）
+# 可选模型: YOLOv8n, YOLOv8s, YOLOv8m, YOLOv8l, YOLOv8x (nano, small, medium, large, xlarge)
+model = YOLO('yolov8n.pt')
+
+# 2. 训练模型
+results = model.train(
+    data='coco128.yaml',  # 数据集配置文件，YOLOv8会自动识别并下载
+    epochs=50,            # 训练轮次，建议50-100轮以观察效果
+    imgsz=640,            # 输入图像大小
+    batch=16,             # 批次大小，根据GPU内存调整（CPU可设为-1）
+    name='yolov8n_coco128_exp1', # 实验名称，用于保存结果
+    device='cpu'          # 使用CPU训练，如果有GPU且环境配置好，可改为 device=0 或 'cuda'
+)

Lab-4/src/val.py

@@ -0,0 +1,14 @@
+from ultralytics import YOLO
+
+# 加载训练得到的最佳模型
+best_model = YOLO('runs/detect/yolov8n_coco128_exp1/weights/best.pt')
+# 在验证集上评估模型
+metrics = best_model.val()
+# 程序会自动打印出mAP、 Precision、 Recall等关键指标
+
+# 加载训练得到的最佳模型
+best_model = YOLO('runs/detect/yolov8n_coco128_exp1/weights/best.pt')
+
+# 在验证集上评估模型
+metrics = best_model.val()
+# 程序会自动打印出mAP、 Precision、 Recall等关键指标。

Lab-5/src/detect.py

@@ -0,0 +1,20 @@
+from ultralytics import YOLO
+import cv2
+
+# 加载最佳模型
+model = YOLO('runs/detect/yolov8n_pcb_defect_exp14/weights/best.pt')
+
+# 在单张图像上推理
+results = model('../dataset/images/val/patches_296.jpg', save=True, conf=0.5)
+
+# 可视化结果（使用OpenCV）
+for result in results:
+    plotted_img = result.plot()  # 绘制检测结果
+    cv2.imshow('Detection Result', plotted_img)
+    cv2.waitKey(0)
+    cv2.destroyAllWindows()
+
+# 支持文件夹、视频、摄像头推理
+# results = model('path/to/video.mp4', save=True, show=True)  # 视频文件
+# results = model(0, show=True)  # 摄像头（0为默认摄像头）
+

Lab-5/src/pcb_defect.yaml

@@ -0,0 +1,8 @@
+path: ../dataset  # 数据集根目录
+train: images/train      # 训练图像路径
+val: images/val          # 验证图像路径
+
+# 类别数
+nc: 6
+# 类别名称
+names: ['missing_hole', 'mouse_bite', 'open_circuit', 'short', 'spur', 'spurious_copper']

Lab-5/src/train.py

@@ -0,0 +1,16 @@
+from ultralytics import YOLO
+
+# 加载预训练模型
+model = YOLO('yolov8n.pt')  # 可选模型: yolov8n, yolov8s, yolov8m, yolov8l, yolov8x
+
+# 训练模型
+results = model.train(
+    data='pcb_defect.yaml',  # 数据集配置文件路径
+    epochs=20,              # 训练轮次
+    imgsz=640,               # 输入图像大小
+    batch=16,                # 批次大小（根据GPU内存调整）
+    name='yolov8n_pcb_defect_exp1',  # 实验名称
+    optimizer='AdamW',       # 优化器
+    lr0=0.001,               # 初始学习率
+    device='cpu'             # 使用设备（如GPU：device=0）
+)

Lab-5/src/transform_yolo.py

@@ -0,0 +1,61 @@
+import xml.etree.ElementTree as ET
+
+import pickle
+import os
+from os import listdir, getcwd
+from os.path import join
+import glob
+
+classes = ['crazing', 'inclusion', 'patches', 'pitted_surface', 'rolled-in_scale', 'scratches']  # xml文件中标记的种类
+
+
+def convert(size, box):
+    dw = 1.0 / size[0]
+    dh = 1.0 / size[1]
+    x = (box[0] + box[1]) / 2.0
+    y = (box[2] + box[3]) / 2.0
+    w = box[1] - box[0]
+    h = box[3] - box[2]
+    x = x * dw
+    w = w * dw
+    y = y * dh
+    h = h * dh
+    return (x, y, w, h)
+
+
+def convert_annotation(image_name):
+    try:
+        in_file = open('../dataset/annotations/val/' + image_name[:-3] + 'xml', encoding='utf-8')  # 原来的xml文件路径
+    except FileNotFoundError:
+        print(f"Warning: XML file for {image_name} not found, skipping.")
+        return
+
+    out_file = open('../dataset/labels/val/' + image_name[:-3] + 'txt', 'w', encoding='utf-8')  # 转换后的txt文件存放路径
+    tree = ET.parse(in_file)
+    root = tree.getroot()
+    size = root.find('size')
+    w = int(size.find('width').text)
+    h = int(size.find('height').text)
+
+    for obj in root.iter('object'):
+        difficult = obj.find('difficult').text
+        cls = obj.find('name').text
+        if cls not in classes or int(difficult) == 1:
+            continue
+        cls_id = classes.index(cls)
+        xmlbox = obj.find('bndbox')
+        b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text),
+             float(xmlbox.find('ymax').text))
+        bb = convert((w, h), b)
+        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')
+    in_file.close()
+    out_file.close()
+
+
+wd = getcwd()
+
+if __name__ == '__main__':
+
+    for image_path in glob.glob("../dataset/images/val/*.jpg"):  # xml对应的图片的路径
+        image_name = image_path.split('\\')[-1]
+        convert_annotation(image_name)

Lab-5/src/val.py

@@ -0,0 +1,9 @@
+from ultralytics import YOLO
+
+# 加载最佳模型
+best_model = YOLO('runs/detect/yolov8n_pcb_defect_exp14/weights/best.pt')
+
+# 在测试集上评估
+metrics = best_model.val()  # 程序会自动计算mAP等指标
+print(f"mAP@0.5: {metrics.box.map50}")
+print(f"mAP@0.5:0.95: {metrics.box.map}")

README.md

@@ -22,4 +22,36 @@
 
 ### RGB三通道分离并显示。
 
-`Lab-1/src/rgb_img.py`
+`Lab-1/src/rgb_img.py`
+
+## **实验二** 基于PyTorch的Minist数据集分类
+
+### 数据预览测试和数据装载；
+
+`Lab-2/src/data_load.py`
+
+### 进行数据处理变换操作；
+
+`Lab-2/src/data_transform.py`
+
+### 模型搭建和参数优化；
+
+`Lab-2/src/model_training.py`
+
+### 基于pytorch卷积神经网络的模型训练；
+
+`Lab-2/src/model_training.py`
+
+### 基于pytorch卷积神经网络的模型测试。
+
+`Lab-2/src/model_training.py`
+
+## **实验三** 迁移resnet50猫狗图像分类
+
+### 通过代码自动下载模型并直接调用；对当前迁移过来的模型进行全连接层的调整；模型训练及测试结果。
+
+`Lab-3/src/train.py`
+
+## **实验四** 基于YOLO的目标检测
+
+## **实验五** 基于YOLOv8的工业缺陷检测实验