验证yolo数据集合是否正确

最新推荐文章于 2025-03-28 22:25:05 发布

m0_46308537

最新推荐文章于 2025-03-28 22:25:05 发布

阅读量992

点赞数

分类专栏：深度学习YOLOV5 文章标签： opencv python 计算机视觉

本文链接：https://blog.youkuaiyun.com/m0_46308537/article/details/128112306

版权

Python 图像处理边界框类别标签文件读取

关键词由优快云通过智能技术生成

深度学习YOLOV5 专栏收录该内容

2 篇文章

订阅专栏

这段代码展示了如何读取类别名称文件，解析图像标注文件，并在图像上绘制边界框和类别标签。它涉及到图像处理、文件操作和对象检测的基础步骤。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

def read_class_names():
    """
    loads class name from a file
    加载id和类别名称映射关系，文件中是每行一个类别名称
    """
    names = {}
    names2 = {}
    with open(r'.\Data3\classes_xinpian.txt', 'r') as data:
        for ID, name in enumerate(data):
            names[name.strip('\n')] = ID
            names2[ID]=name.strip('\n')
    print(names2)
    return names,names2

import  os
import cv2
def test_yolo():
    for i in os.listdir(r".\Data3\labels"):
        text_path=os.path.join(r".\Data3\labels",i)
        image_path=text_path.replace("labels","PIC").replace("txt","bmp")
        # image_path=r".\Data3\PIC\123456.bmp"
        # img = cv2.imread(image_path)
        print(image_path)
        img = cv2.imread(image_path)
        img_height, img_width = cv2.imread(image_path).shape[0], cv2.imread(image_path).shape[1]
        print(img_height, img_width)
        _,names2=read_class_names()
        with open(text_path, "r") as fd:
            datas = fd.readlines()

            for data in datas:

                if len(data) > 1:
                    data_list=data.strip('\n').split(' ')
                    center_x = round(float(data_list[1]) * img_width)
                    center_y = round(float(data_list[2]) * img_height)
                    bbox_width = round(float(data_list[3]) * img_width)
                    bbox_height = round(float(data_list[4]) * img_height)
                    x_min = str(int(center_x - bbox_width / 2))
                    y_min = str(int(center_y - bbox_height / 2))
                    x_max = str(int(center_x + bbox_width / 2))
                    y_max = str(int(center_y + bbox_height / 2))

                    label =names2[int(data_list[0])]
                    print(label)
                    print(x_min,y_min,x_max,y_max,label)
                #
                    cv2.putText(img,label,(int(x_min),int(y_min)-2),0,1,[255,255,255])
                    cv2.rectangle(img, (int(x_min), int(y_min)), (int(x_max), int(y_max)), (255, 0, 0), 2)
        cv2.namedWindow('img', 0)
        cv2.resizeWindow('img', int(1920), int(1680))
        cv2.imshow('img', img)
        cv2.waitKey(0)
        cv2.destroyAllWindows()

test_yolo()
    # with open(,"r") as fd: