YOLO数据集增强、扩充_yolo数据集扩充

YOLO数据集增强、扩充

介绍

yolo已标注数据集增强扩充，已标注的.txt格式的数据集增强，包含旋转、平移、裁剪、翻转、调整亮度和增加噪声6种方式，随意组合，每张图片增强至5张。

核心代码

translation_range = np.random.randint(10, 50)  
rotation_angle_range = np.random.randint(10, 30)  
crop_range = np.random.randint(10, 30)  
brightness_range = (np.random.uniform(0.5, 1.5), np.random.uniform(1.5, 2.5))  
noise_stddev = np.random.randint(5, 20)  

# 获取图像文件和标注文件的路径
img_paths = [os.path.join(src_img_dir, f) for f in os.listdir(src_img_dir) if f.endswith(('.jpg', '.jpeg', '.png'))]
txt_paths = [os.path.join(src_txt_dir, f) for f in os.listdir(src_txt_dir) if f.endswith('.txt')]

for img_path, txt_path in zip(img_paths, txt_paths):
    # 读取图像
    image = cv2.imread(img_path)
    h, w, _ = image.shape

    # 读取标注数据
    with open(txt_path, 'r') as txt_file:
        lines = txt_file.readlines()

    augmentations = np.random.choice(['mirror', 'translate', 'rotate', 'brightness', 'noise', 'crop', 'flip'], size=5, replace=False)

    for i, aug in enumerate(augmentations):
        image_augmented = image.copy()
        lines_augmented = list(lines)

        if aug == 'mirror':
            image_augmented = cv2.flip(image_augmented, 1)
            for j, line in enumerate(lines_augmented):
                values = line.strip().split(' ')
                if len(values) == 5:
                    x, y, width, height = map(float, values[1:5])
                    x = 1 - x
                    lines_augmented[j] = f"{values[0]} {x} {values[2]} {values[3]} {values[4]}\n"

        if aug == 'translate':
            tx = np.random.randint(-translation_range, translation_range)
            ty = np.random.randint(-translation_range, translation_range)
            M = np.float32([[1, 0, tx], [0, 1, ty]])
            image_augmented = cv2.warpAffine(image_augmented, M, (w, h))

            for j, line in enumerate(lines_augmented):
                values = line.strip().split(' ')
                if len(values) == 5:
                    x, y, width, height = map(float, values[1:5])
                    x += tx / w
                    y += ty / h
                    lines_augmented[j] = f"{values[0]} {x} {y} {values[2]} {values[3]}\n"

        if aug == 'rotate':
            angle = np.random.randint(-rotation_angle_range, rotation_angle_range)
            M = cv2.getRotationMatrix2D((w / 2, h / 2), angle, 1)
            image_augmented = cv2.warpAffine(image_augmented, M, (w, h))

            for j, line in enumerate(lines_augmented):
                values = line.strip().split(' ')
                if len(values) == 5:
                    x, y, width, height = map(float, values[1:5])
                    x_rot, y_rot = np.dot(M, np.array([x * w, y * h, 1]))
                    x_rot /= w
                    y_rot /= h
                    lines_augmented[j] = f"{values[0]} {x_rot} {y_rot} {values[2]} {values[3]}\n"

        if aug == 'brightness':
            brightness_factor = np.random.uniform(*brightness_range)
            image_augmented = cv2.convertScaleAbs(image_augmented, alpha=brightness_factor, beta=0)

        if aug == 'noise':
            noise = np.random.normal(0, noise_stddev, image_augmented.shape).astype(np.uint8)
            image_augmented = cv2.add(image_augmented, noise)

        if aug == 'crop':
            x1 = np.random.randint(0, crop_range)
            y1 = np.random.randint(0, crop_range)
            x2 = w - np.random.randint(0, crop_range)
            y2 = h - np.random.randint(0, crop_range)
            image_augmented = image_augmented[y1:y2, x1:x2]
            for j, line in enumerate(lines_augmented):
                values = line.strip().split(' ')
                if len(values) == 5:
                    x, y, width, height = map(float, values[1:5])
                    x = (x - x1) / (x2 - x1)
                    y = (y - y1) / (y2 - y1)
                    lines_augmented[j] = f"{values[0]} {x} {y} {values[2]} {values[3]}\n"

        if aug == 'flip':
            image_augmented = cv2.flip(image_augmented, 0)
            for j, line in enumerate(lines_augmented):
                values = line.strip().split(' ')
                if len(values) == 5:
                    y = 1 - float(values[2])
                    lines_augmented[j] = f"{values[0]} {values[1]} {y} {values[2]} {values[3]}\n"

        # 保存增强后的图像
        img_name = os.path.basename(img_path)
        output_img_path = os.path.join(output_img_dir, f"{i}_{os.path.splitext(img_name)[0]}.png")
        cv2.imwrite(output_img_path, image_augmented)

        # 保存增强后的标注数据
        txt_name = os.path.basename(txt_path)
        output_txt_path = os.path.join(output_txt_dir, f"{i}_{txt_name}")
        with open(output_txt_path, 'w') as output_txt_file:
            output_txt_file.writelines(lines_augmented)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102

注意点

• opencv-python的版本不能过高，否则报错
• 本例中的版本为4.1.2.30

完整代码

• 连接：https://github.com/384520505/YoloDatasetsEnhance