bev-project/tools/visualize_vector_map.py

"""
可视化矢量地图标注

用于验证提取的矢量地图数据是否正确
"""

import pickle
import argparse
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.patches import Polygon as MPLPolygon
from matplotlib.collections import LineCollection


def visualize_vector_map(vector_map_file, num_samples=5, save_dir='vis_vector_map'):
    """
    可视化矢量地图
    
    Args:
        vector_map_file: 矢量地图pkl文件
        num_samples: 可视化的样本数
        save_dir: 保存目录
    """
    import os
    os.makedirs(save_dir, exist_ok=True)
    
    # 加载数据
    print(f"加载矢量地图: {vector_map_file}")
    with open(vector_map_file, 'rb') as f:
        vector_maps = pickle.load(f)
    
    print(f"总样本数: {len(vector_maps)}")
    
    # 随机选择几个样本
    sample_tokens = list(vector_maps.keys())[:num_samples]
    
    for idx, token in enumerate(sample_tokens):
        vectors = vector_maps[token]
        
        # 创建图
        fig, ax = plt.subplots(figsize=(12, 12))
        ax.set_xlim(0, 1)
        ax.set_ylim(0, 1)
        ax.set_aspect('equal')
        ax.grid(True, alpha=0.3)
        ax.set_title(f'Sample {idx+1}: {token[:8]}...')
        ax.set_xlabel('X (normalized)')
        ax.set_ylabel('Y (normalized)')
        
        # 绘制divider（蓝色）
        for vec in vectors.get('divider', []):
            points = np.array(vec['points'])
            if len(points) > 0:
                ax.plot(points[:, 0], points[:, 1], 'b-', linewidth=2, alpha=0.7, label='Divider')
                ax.scatter(points[:, 0], points[:, 1], c='b', s=10, alpha=0.5)
        
        # 绘制boundary（红色）
        for vec in vectors.get('boundary', []):
            points = np.array(vec['points'])
            if len(points) > 0:
                ax.plot(points[:, 0], points[:, 1], 'r-', linewidth=2, alpha=0.7, label='Boundary')
                ax.scatter(points[:, 0], points[:, 1], c='r', s=10, alpha=0.5)
        
        # 绘制ped_crossing（绿色）
        for vec in vectors.get('ped_crossing', []):
            points = np.array(vec['points'])
            if len(points) > 0:
                ax.plot(points[:, 0], points[:, 1], 'g-', linewidth=2, alpha=0.7, label='Ped Crossing')
                ax.scatter(points[:, 0], points[:, 1], c='g', s=10, alpha=0.5)
        
        # 去重图例
        handles, labels = ax.get_legend_handles_labels()
        by_label = dict(zip(labels, handles))
        ax.legend(by_label.values(), by_label.keys(), loc='upper right')
        
        # 统计
        num_dividers = len(vectors.get('divider', []))
        num_boundaries = len(vectors.get('boundary', []))
        num_crossings = len(vectors.get('ped_crossing', []))
        
        ax.text(0.02, 0.98, 
                f'Dividers: {num_dividers}\nBoundaries: {num_boundaries}\nCrossings: {num_crossings}',
                transform=ax.transAxes, 
                verticalalignment='top',
                bbox=dict(boxstyle='round', facecolor='wheat', alpha=0.5))
        
        # 保存
        save_path = f'{save_dir}/sample_{idx+1:03d}.png'
        plt.savefig(save_path, dpi=150, bbox_inches='tight')
        print(f"保存到: {save_path}")
        plt.close()
    
    print(f"\n✅ 可视化完成！图像保存在: {save_dir}/")


def print_vector_statistics(vector_map_file):
    """打印矢量地图统计信息"""
    with open(vector_map_file, 'rb') as f:
        vector_maps = pickle.load(f)
    
    total_dividers = 0
    total_boundaries = 0
    total_crossings = 0
    
    divider_points = []
    boundary_points = []
    crossing_points = []
    
    for vectors in vector_maps.values():
        dividers = vectors.get('divider', [])
        boundaries = vectors.get('boundary', [])
        crossings = vectors.get('ped_crossing', [])
        
        total_dividers += len(dividers)
        total_boundaries += len(boundaries)
        total_crossings += len(crossings)
        
        for vec in dividers:
            divider_points.append(len(vec['points']))
        for vec in boundaries:
            boundary_points.append(len(vec['points']))
        for vec in crossings:
            crossing_points.append(len(vec['points']))
    
    print("\n========== 矢量地图统计 ==========")
    print(f"总样本数: {len(vector_maps)}")
    print(f"\nDividers:")
    print(f"  总数: {total_dividers}")
    print(f"  平均/样本: {total_dividers/len(vector_maps):.2f}")
    if divider_points:
        print(f"  平均点数: {np.mean(divider_points):.1f}")
    
    print(f"\nBoundaries:")
    print(f"  总数: {total_boundaries}")
    print(f"  平均/样本: {total_boundaries/len(vector_maps):.2f}")
    if boundary_points:
        print(f"  平均点数: {np.mean(boundary_points):.1f}")
    
    print(f"\nPed Crossings:")
    print(f"  总数: {total_crossings}")
    print(f"  平均/样本: {total_crossings/len(vector_maps):.2f}")
    if crossing_points:
        print(f"  平均点数: {np.mean(crossing_points):.1f}")
    print("================================\n")


def main():
    parser = argparse.ArgumentParser(description='可视化矢量地图')
    parser.add_argument('--vector-map-file', type=str,
                       default='data/nuscenes/vector_maps_bevfusion.pkl',
                       help='矢量地图文件')
    parser.add_argument('--num-samples', type=int, default=5,
                       help='可视化的样本数')
    parser.add_argument('--save-dir', type=str, default='vis_vector_map',
                       help='保存目录')
    parser.add_argument('--stats-only', action='store_true',
                       help='只打印统计信息')
    
    args = parser.parse_args()
    
    # 打印统计
    print_vector_statistics(args.vector_map_file)
    
    # 可视化
    if not args.stats_only:
        visualize_vector_map(
            args.vector_map_file,
            num_samples=args.num_samples,
            save_dir=args.save_dir
        )


if __name__ == '__main__':
    main()