SemanticKITTIAPI/generate_test_data.py at main · yankmo/SemanticKITTIAPI · GitHub

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
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
#!/usr/bin/env python3
"""
生成测试数据脚本
为演示和测试目的创建示例点云数据
"""

import numpy as np
import os
import argparse

def generate_test_pointcloud(n_points=10000, scene_type="urban"):
    """生成测试点云数据"""

    if scene_type == "urban":
        # 城市场景：道路、建筑、车辆
        points_per_class = n_points // 3

        # 道路 (平面，z接近0)
        road_points = np.random.uniform(-50, 50, (points_per_class, 2))
        road_z = np.random.normal(-1.5, 0.2, points_per_class)
        road_xyz = np.column_stack([road_points, road_z])
        road_labels = np.full(points_per_class, 9)  # road class

        # 建筑物 (高结构)
        building_x = np.random.uniform(-30, 30, points_per_class)
        building_y = np.random.uniform(-30, 30, points_per_class)
        building_z = np.random.uniform(0, 15, points_per_class)
        building_xyz = np.column_stack([building_x, building_y, building_z])
        building_labels = np.full(points_per_class, 13)  # building class

        # 车辆 (中等高度) - 补齐剩余点数
        remaining_points = n_points - 2 * points_per_class
        vehicle_x = np.random.uniform(-20, 20, remaining_points)
        vehicle_y = np.random.uniform(-20, 20, remaining_points)
        vehicle_z = np.random.uniform(-1, 2, remaining_points)
        vehicle_xyz = np.column_stack([vehicle_x, vehicle_y, vehicle_z])
        vehicle_labels = np.full(remaining_points, 1)  # car class

        points = np.vstack([road_xyz, building_xyz, vehicle_xyz])
        labels = np.hstack([road_labels, building_labels, vehicle_labels])

    elif scene_type == "highway":
        # 高速公路场景
        points_per_class = n_points // 2

        # 道路中心线
        road_center = np.linspace(-100, 100, points_per_class)
        road_y = np.random.normal(0, 5, points_per_class)
        road_z = np.random.normal(-1.5, 0.1, points_per_class)
        road_xyz = np.column_stack([road_center, road_y, road_z])
        road_labels = np.full(points_per_class, 9)

        # 植被 (道路两侧) - 补齐剩余点数
        remaining_points = n_points - points_per_class
        veg_x = np.random.uniform(-100, 100, remaining_points)
        veg_y = np.random.choice([-20, 20], remaining_points) + np.random.normal(0, 5, remaining_points)
        veg_z = np.random.uniform(0, 10, remaining_points)
        veg_xyz = np.column_stack([veg_x, veg_y, veg_z])
        veg_labels = np.full(remaining_points, 15)  # vegetation

        points = np.vstack([road_xyz, veg_xyz])
        labels = np.hstack([road_labels, veg_labels])

    else:  # random
        points = np.random.randn(n_points, 3) * 20
        labels = np.random.randint(0, 20, n_points)

    # 添加强度信息 - 确保与点数匹配
    intensity = np.random.uniform(0, 1, len(points))

    return points, intensity, labels

def save_kitti_format(points, intensity, labels, scan_path, label_path):
    """保存为KITTI格式"""
    # 保存点云 (.bin格式)
    scan_data = np.column_stack([points, intensity]).astype(np.float32)
    scan_data.tofile(scan_path)

    # 保存标签 (.label格式)
    labels.astype(np.uint32).tofile(label_path)

def create_test_dataset(output_dir="test_data", n_scans=5, n_points=10000):
    """创建完整的测试数据集"""

    # 创建目录结构
    sequences_dir = os.path.join(output_dir, "sequences", "00")
    velodyne_dir = os.path.join(sequences_dir, "velodyne")
    labels_dir = os.path.join(sequences_dir, "labels")

    os.makedirs(velodyne_dir, exist_ok=True)
    os.makedirs(labels_dir, exist_ok=True)

    scene_types = ["urban", "highway", "random"]

    print(f"生成 {n_scans} 个测试扫描文件...")

    for i in range(n_scans):
        # 选择场景类型
        scene_type = scene_types[i % len(scene_types)]

        # 生成数据
        points, intensity, labels = generate_test_pointcloud(n_points, scene_type)

        # 文件名
        scan_file = f"{i:06d}.bin"
        label_file = f"{i:06d}.label"

        scan_path = os.path.join(velodyne_dir, scan_file)
        label_path = os.path.join(labels_dir, label_file)

        # 保存文件
        save_kitti_format(points, intensity, labels, scan_path, label_path)

        print(f"  ✓ 生成扫描 {i:06d} ({scene_type} 场景)")

    # 创建不同模型的预测目录
    models = ["model_a", "model_b", "model_c"]

    for model_name in models:
        model_pred_dir = os.path.join(sequences_dir, f"predictions_{model_name}")
        os.makedirs(model_pred_dir, exist_ok=True)

        print(f"为 {model_name} 生成预测结果...")

        for i in range(n_scans):
            # 加载原始点云以获取点数
            scan_path = os.path.join(velodyne_dir, f"{i:06d}.bin")
            scan_data = np.fromfile(scan_path, dtype=np.float32).reshape(-1, 4)
            n_points_actual = scan_data.shape[0]

            # 生成不同的预测策略
            if model_name == "model_a":
                # 基于距离的预测
                points = scan_data[:, :3]
                distances = np.linalg.norm(points, axis=1)
                predictions = np.zeros(n_points_actual, dtype=np.uint32)
                predictions[distances < 10] = 1   # 近距离 - 车辆
                predictions[(distances >= 10) & (distances < 30)] = 9  # 中距离 - 道路
                predictions[distances >= 30] = 15  # 远距离 - 植被

            elif model_name == "model_b":
                # 基于高度的预测
                heights = scan_data[:, 2]
                predictions = np.zeros(n_points_actual, dtype=np.uint32)
                predictions[heights < -1] = 9     # 低处 - 道路
                predictions[(heights >= -1) & (heights < 1)] = 1   # 车辆高度
                predictions[(heights >= 1) & (heights < 5)] = 6    # 人员高度
                predictions[heights >= 5] = 13   # 建筑物

            else:  # model_c
                # 随机预测（模拟较差的模型）
                predictions = np.random.randint(0, 20, n_points_actual, dtype=np.uint32)

            # 保存预测结果
            pred_path = os.path.join(model_pred_dir, f"{i:06d}.label")
            predictions.tofile(pred_path)

        print(f"  ✓ {model_name} 预测结果已生成")

    print(f"\n✓ 测试数据集创建完成: {output_dir}")
    print(f"  - {n_scans} 个扫描文件")
    print(f"  - {len(models)} 个模型的预测结果")
    print(f"  - 每个扫描约 {n_points:,} 个点")

def main():
    parser = argparse.ArgumentParser(description="生成测试点云数据")
    parser.add_argument('--output', '-o', type=str, default='test_data',
                       help='输出目录 (默认: test_data)')
    parser.add_argument('--scans', '-s', type=int, default=5,
                       help='生成的扫描文件数量 (默认: 5)')
    parser.add_argument('--points', '-p', type=int, default=10000,
                       help='每个扫描的点数 (默认: 10000)')

    args = parser.parse_args()

    create_test_dataset(args.output, args.scans, args.points)

if __name__ == "__main__":
    main()