根据Results.csv文件画性能曲线图

import pandas as pd
import matplotlib.pyplot as plt
from scipy.ndimage import gaussian_filter1d
import os
import math

# 训练结果列表 (请确保这里的路径是正确的)
# Training results file list (Please ensure the paths here are correct)
results_files = [
    r'F:\ultralyticsPro\yolocsv\ours.csv',
    r'F:\ultralyticsPro\yolocsv\yolov5.csv',
    r'F:\ultralyticsPro\yolocsv\yolov8.csv',
    r'F:\ultralyticsPro\yolocsv\yolov9.csv',
    r'F:\ultralyticsPro\yolocsv\yolov10.csv',
    r'F:\ultralyticsPro\yolocsv\yolov11.csv',
    r'F:\ultralyticsPro\yolocsv\yolov12.csv',
]

# 与results_files顺序对应的自定义标签
# Custom labels corresponding to the order of results_files
custom_labels = [
    'ours',
    'YOLOv5n',
    'YOLOv8n',
    'YOLOv9n',
    'YOLOv10n',
    'YOLOv11n',
    'YOLOv12n',
]

# 定义不同的线型和标记点用于区分曲线
# Define different linestyles and markers to distinguish curves
linestyles = ['-', '--', '-.', ':'] # 实线, 虚线, 点划线, 点线 Solid, dashed, dash-dot, dotted
markers = ['o', 's', '^', 'v', 'D', '*', 'p'] # 圆, 方块, 上三角, 下三角, 菱形, 星号, 五边形 Circle, square, triangle_up, triangle_down, diamond, star, pentagon

def plot_metrics_grid(
        metrics_row1,
        labels_row1,
        metrics_row2,
        labels_row2,
        custom_labels,
        results_files,
        output_path='metrics_grid_plot.png',
        max_x_axis_epoch=200,
        sigma=1,
        dpi=300
):
    """
    绘制性能指标的网格图，分为两行显示，每个指标一个独立的坐标框。
    使用不同的颜色、线型和标记点区分不同模型的曲线。
    网格布局会根据指标数量动态调整。

    Draws a grid plot of performance metrics, split into two rows,
    with each metric in its own subplot.
    Uses different colors, linestyles, and markers to distinguish curves for different models.
    The grid layout adjusts dynamically based on the number of metrics per row.

    Args:
        metrics_row1 (list): 第 1 行性能指标的键名列表.
        labels_row1 (list): 第 1 行性能指标的显示标签列表.
        metrics_row2 (list): 第 2 行性能指标的键名列表.
        labels_row2 (list): 第 2 行性能指标的显示标签列表.
        custom_labels (list): 每个结果文件的自定义标签.
        results_files (list): CSV 文件路径列表.
        output_path (str): 输出图像的文件路径.
        max_x_axis_epoch (int): X 轴（Epoch）的最大值.
        sigma (float): 高斯平滑的 sigma 值.
        dpi (int): 保存图像的分辨率.
    """
    num_metrics_r1 = len(metrics_row1)
    num_metrics_r2 = len(metrics_row2)
    ncols = max(num_metrics_r1, num_metrics_r2)

    # 创建图形和 GridSpec (2 行) - 增大 figsize 使子图更大
    # Create figure and GridSpec (2 rows) - Increase figsize for larger subplots
    fig = plt.figure(figsize=(max(14, 7 * ncols), 14)) # 增大图形尺寸 Increase figure size
    gs = fig.add_gridspec(2, ncols)

    # --- 绘制第 1 行性能指标 ---
    print("--- 开始绘制第 1 行性能指标 ---")
    for i, (metric_key, metric_label) in enumerate(zip(metrics_row1, labels_row1)):
        ax = fig.add_subplot(gs[0, i])
        print(f"绘制指标: {metric_label} ({metric_key})")
        plot_count = 0
        for idx, (file_path, custom_label) in enumerate(zip(results_files, custom_labels)):
            try:
                if not os.path.exists(file_path):
                    print(f"警告: 文件未找到，跳过: {file_path}")
                    continue
                df = pd.read_csv(file_path)
                df.columns = df.columns.str.strip()
                if df.empty:
                    print(f"警告: 文件为空或读取失败，跳过: {file_path}")
                    continue
                if 'epoch' not in df.columns or metric_key not in df.columns:
                    print(f"警告: 文件 {os.path.basename(file_path)} 缺少列 'epoch' 或 '{metric_key}'，跳过。")
                    continue
                if df[metric_key].isnull().all() or not pd.api.types.is_numeric_dtype(df[metric_key]):
                     print(f"警告: 文件 {os.path.basename(file_path)} 中的列 '{metric_key}' 数据无效或非数值，跳过。")
                     continue

                epochs_to_plot = df['epoch']
                y_data = df[metric_key].astype(float)
                y_smooth = gaussian_filter1d(y_data, sigma=sigma)
                max_epoch_data = epochs_to_plot.max()

                # 应用不同的线型和标记点
                current_linestyle = linestyles[idx % len(linestyles)]
                current_marker = markers[idx % len(markers)]
                line, = ax.plot(epochs_to_plot, y_smooth, label=f'{custom_label}',
                                linewidth=2.5,
                                linestyle=current_linestyle,
                                marker=current_marker,
                                markersize=4,
                                markevery=10
                               )
                plot_count += 1
                print(f"  已绘制: {custom_label} (来自 {os.path.basename(file_path)}, 线型: {current_linestyle}, 标记: {current_marker})")

                if max_epoch_data < max_x_axis_epoch:
                    last_y_value = y_smooth[-1]
                    last_epoch_value = epochs_to_plot.iloc[-1]
                    ax.plot(
                        [last_epoch_value, last_epoch_value],
                        [last_y_value, 0],
                        linestyle='dotted',
                        color=line.get_color(),
                    )
            except FileNotFoundError:
                print(f"错误: 文件未找到: {file_path}")
            except pd.errors.EmptyDataError:
                print(f"错误: 文件为空: {file_path}")
            except Exception as e:
                print(f"错误: 读取或处理文件 {os.path.basename(file_path)} 时发生错误: {e}")

        ax.set_title(f'{metric_label}', fontsize=14)
        ax.set_xlabel('Epochs', fontsize=12)
        ax.set_ylabel('Value', fontsize=12)
        if plot_count > 0:
             ax.legend(fontsize=10, loc='best')
        ax.set_ylim(bottom=0)
        ax.set_xlim(left=0, right=max_x_axis_epoch)
        ax.grid(True)

    # --- 绘制第 2 行性能指标 ---
    print("\n--- 开始绘制第 2 行性能指标 ---")
    for i, (metric_key, metric_label) in enumerate(zip(metrics_row2, labels_row2)):
        ax = fig.add_subplot(gs[1, i])
        print(f"绘制指标: {metric_label} ({metric_key})")
        plot_count = 0
        for idx, (file_path, custom_label) in enumerate(zip(results_files, custom_labels)):
            try:
                if not os.path.exists(file_path):
                    print(f"警告: 文件未找到，跳过: {file_path}")
                    continue
                df = pd.read_csv(file_path)
                df.columns = df.columns.str.strip()
                if df.empty:
                    print(f"警告: 文件为空或读取失败，跳过: {file_path}")
                    continue
                if 'epoch' not in df.columns or metric_key not in df.columns:
                    print(f"警告: 文件 {os.path.basename(file_path)} 缺少列 'epoch' 或 '{metric_key}'，跳过。")
                    continue
                if df[metric_key].isnull().all() or not pd.api.types.is_numeric_dtype(df[metric_key]):
                     print(f"警告: 文件 {os.path.basename(file_path)} 中的列 '{metric_key}' 数据无效或非数值，跳过。")
                     continue

                epochs_to_plot = df['epoch']
                y_data = df[metric_key].astype(float)
                y_smooth = gaussian_filter1d(y_data, sigma=sigma)
                max_epoch_data = epochs_to_plot.max()

                # 应用不同的线型和标记点
                current_linestyle = linestyles[idx % len(linestyles)]
                current_marker = markers[idx % len(markers)]
                line, = ax.plot(epochs_to_plot, y_smooth, label=f'{custom_label}',
                                linewidth=2.5,
                                linestyle=current_linestyle,
                                marker=current_marker,
                                markersize=4,
                                markevery=10
                               )
                plot_count += 1
                print(f"  已绘制: {custom_label} (来自 {os.path.basename(file_path)}, 线型: {current_linestyle}, 标记: {current_marker})")

                if max_epoch_data < max_x_axis_epoch:
                    last_y_value = y_smooth[-1]
                    last_epoch_value = epochs_to_plot.iloc[-1]
                    ax.plot(
                        [last_epoch_value, last_epoch_value],
                        [last_y_value, 0],
                        linestyle='dotted',
                        color=line.get_color(),
                    )
            except FileNotFoundError:
                print(f"错误: 文件未找到: {file_path}")
            except pd.errors.EmptyDataError:
                print(f"错误: 文件为空: {file_path}")
            except Exception as e:
                print(f"错误: 读取或处理文件 {os.path.basename(file_path)} 时发生错误: {e}")

        ax.set_title(f'{metric_label}', fontsize=14)
        ax.set_xlabel('Epochs', fontsize=12)
        ax.set_ylabel('Value', fontsize=12)
        if plot_count > 0:
             ax.legend(fontsize=10, loc='best')
        ax.set_ylim(bottom=0)
        ax.set_xlim(left=0, right=max_x_axis_epoch)
        ax.grid(True)

    # 调整布局并保存图像
    plt.tight_layout(pad=2.0, h_pad=3.0)
    if output_path:
        try:
            output_dir = os.path.dirname(output_path)
            if output_dir:
                 os.makedirs(output_dir, exist_ok=True)
            plt.savefig(output_path, dpi=dpi, bbox_inches='tight')
            print(f"\n图像已成功保存到: {output_path}, dpi={dpi}")
        except Exception as e:
            print(f"\n错误: 保存图像到 {output_path} 时失败: {e}")
    else:
        print("\n警告: output_path 为空或无效，图像未保存。")

    # plt.show()


if __name__ == '__main__':
    metrics_row1 = [
        'metrics/precision(B)',
        'metrics/recall(B)',
    ]
    labels_row1 = ['Precision', 'Recall']

    metrics_row2 = [
        'metrics/mAP50(B)',
        'metrics/mAP50-95(B)',
    ]
    labels_row2 = ['mAP@50', 'mAP@50-95']

    output_path_grid = r'F:\ultralyticsPro\metrics_grid_plot_perf_markers_larger.png' # 输出路径

    plot_metrics_grid(
        metrics_row1,
        labels_row1,
        metrics_row2,
        labels_row2,
        custom_labels,
        results_files,
        output_path=output_path_grid,
        max_x_axis_epoch=200,
        sigma=1,
        dpi=300
    )