关于YOLOv5训练结果results.txt绘制对比图_yolov5 result图

目录

用两个results.txt在一幅图中画2条对比曲线

用一个results.txt在一幅图中画多条对比曲线：

用两个results.txt在一幅图中画2条对比曲线

# -*- coding:utf-8 -*-
import matplotlib.pyplot as plt
import numpy as np
 
 
column = ['epoch', 'train_GIOU_loss', 'train_obj_loss', 'train_cls_loss', 'total', 'target', 'img_size', 'precision',
          'recall', 'MAP@0.5', 'F1', 'val_GIOU_loss', 'val_obj_loss', 'val_cls_loss']
 
 
def plot_result(result1, result2, title, yl):
    ind = column.index(title)  # 获取索引
    plt.rcParams['font.sans-serif'] = 'Times New Roman'
    plt.figure(figsize=(10, 8), dpi=400)
    x = result1[range(0, 60, 5), 0]
    y1 = result1[range(0, 60, 5), ind]
    y2 = result2[range(0, 60, 5), ind]
    line1 = plt.plot(x, y1, color="#8B4513", linewidth=2, label="yolov5", marker='^')
    line2 = plt.plot(x, y2, color="#90EE90", linewidth=2, label="advanced yolov5", marker='^')
    plt.xticks(range(0, 60, 5), fontsize=14)
    plt.yticks(fontsize=14)
    plt.xlabel('epoch', fontsize=16)
    plt.ylabel(yl, fontsize=16)
    plt.legend(loc="best", fontsize=16, frameon=False)
    plt.savefig('总损失对比图')
    plt.show()
 
 
# 获取results文件中的有效列并且新增epoch序号列
def result2matrix(result_dir, epoch_num):
    data = np.genfromtxt(result_dir)
    result = data[0:epoch_num + 1, 2:15]  # 获取数据
    print(f"\n一共{len(result)}个epoches的结果")
    epoch = np.arange(len(result))  # 生成epoch序号
    epoch = epoch.reshape(len(result), 1)
    result = np.hstack((epoch, result))  # 新添一列epoch
    return result
 
 
def plot_yolov5_curves(txt_dir1, txt_dir2):
    output1 = result2matrix(txt_dir1, epoch_num=60)
    output2 = result2matrix(txt_dir2, epoch_num=60)
    plot_result(output1, output2, 'total', 'total_loss')
 
 
 
'''
一共11个图，可根据自己要求拓展
'''
if __name__ == '__main__':
    dir1 = "results1.txt"
    dir2 = "results2.txt"
    plot_yolov5_curves(dir1, dir2)

用一个results.txt在一幅图中画多条对比曲线：

# -*- coding:utf-8 -*-
import matplotlib.pyplot as plt
import numpy as np
 
 
column = ['epoch', 'train_GIOU_loss', 'train_obj_loss', 'train_cls_loss', 'train_landmark_loss', 'total', 'target', 'img_size', 'precision',
          'recall', 'MAP@0.5', 'F1', 'val_GIOU_loss', 'val_obj_loss', 'val_cls_loss']
 
# 同一幅图中画多条曲线
def plot_result(result1, yl):
    ind1 = column.index('train_GIOU_loss')  # 获取giou索引
    ind2 = column.index('train_obj_loss')  # 获取obj索引索引
    ind3 = column.index('train_cls_loss')  # 获取cls索引
    ind4 = column.index('train_landmark_loss')  # 获取landmark索引
    ind5 = column.index('total')  # 获取total索引
    plt.rcParams['font.sans-serif'] = 'Times New Roman'
    plt.figure(figsize=(10, 8), dpi=400)
    x = result1[range(0, 200, 10), 0]
    # 每条损失曲线对应的y值
    y1 = result1[range(0, 200, 10), ind1]
    y2 = result1[range(0, 200, 10), ind2]
    y3 = result1[range(0, 200, 10), ind3]
    y4 = result1[range(0, 200, 10), ind4]
    y5 = result1[range(0, 200, 10), ind5]
 
    # 画giou
    line1 = plt.plot(x, y1, color="#6F80BE", linewidth=2, label="CIoU_loss", marker='^')
    # 画obj
    line2 = plt.plot(x, y2, color="#67C2A3", linewidth=2, label="obj_loss", marker='8')
    # 画cls
    line3 = plt.plot(x, y3, color="#FC8A61", linewidth=2, label="cls_loss", marker='s')
    # 画landmark
    line4 = plt.plot(x, y4, color="#E889BD", linewidth=2, label="landmark_loss", marker='D')
    # 画total
    line5 = plt.plot(x, y5, color="#C82423", linewidth=2, label="total_loss", marker='>')
 
    plt.xticks(range(0, 200, 10), fontsize=14)
    plt.yticks(fontsize=14)
    plt.xlabel('epoch', fontsize=16)
    plt.ylabel(yl, fontsize=16)
    plt.legend(loc="best", fontsize=16, frameon=False)
    plt.savefig('损失对比图')
    plt.show()
 
 
# 获取results文件中的有效列并且新增epoch序号列
def result2matrix(result_dir, epoch_num):
    data = np.genfromtxt(result_dir)
    result = data[0:epoch_num + 1, 2:16]  # 获取数据
    print(f"\n一共{len(result)}个epoches的结果")
    epoch = np.arange(len(result))  # 生成epoch序号
    epoch = epoch.reshape(len(result), 1)
    result = np.hstack((epoch, result))  # 新添一列epoch
    return result
 
 
def plot_yolov5_curves(txt_dir1):
    output1 = result2matrix(txt_dir1, epoch_num=200)
    plot_result(output1, 'loss')
 
'''
一共11个图，可根据自己要求拓展
'''
if __name__ == '__main__':
    dir1 = "results.txt"
    plot_yolov5_curves(dir1)
 
 

示例如下：