YoloV5实战：手把手教物体检测

matplotlib>=3.2.2 numpy>=1.18.5 opencv-python>=4.1.2 pillow PyYAML>=5.3 scipy>=1.4.1 tensorboard>=2.2 torch>=1.6.0 torchvision>=0.7.0 tqdm>=4.41.0

import os import numpy as np import json from glob import glob import cv2 from sklearn.model_selection import train_test_split from os import getcwd   classes = ["aircraft", "oiltank"] # 1.标签路径 labelme_path = "LabelmeData/" isUseTest = True  # 是否创建test集 # 3.获取待处理文件 files = glob(labelme_path + "*.json") files = [i.replace("\\", "/").split("/")[-1].split(".json")[0] for i in files] print(files) if isUseTest:     trainval_files, test_files = train_test_split(files, test_size=0.1, random_state=55) else:     trainval_files = files # split train_files, val_files = train_test_split(trainval_files, test_size=0.1, random_state=55)     def convert(size, box):     dw = 1. / (size[0])     dh = 1. / (size[1])     x = (box[0] + box[1]) / 2.0 - 1     y = (box[2] + box[3]) / 2.0 - 1     w = box[1] - box[0]     h = box[3] - box[2]     x = x * dw     w = w * dw     y = y * dh     h = h * dh     return (x, y, w, h)     wd = getcwd() print(wd)     def ChangeToYolo5(files, txt_Name):     if not os.path.exists('tmp/'):         os.makedirs('tmp/')     list_file = open('tmp/%s.txt' % (txt_Name), 'w')     for json_file_ in files:         json_filename = labelme_path + json_file_ + ".json"         imagePath = labelme_path + json_file_ + ".jpg"         list_file.write('%s/%s\n' % (wd, imagePath))         out_file = open('%s/%s.txt' % (labelme_path, json_file_), 'w')         json_file = json.load(open(json_filename, "r", encoding="utf-8"))         height, width, channels = cv2.imread(labelme_path + json_file_ + ".jpg").shape         for multi in json_file["shapes"]:             points = np.array(multi["points"])             xmin = min(points[:, 0]) if min(points[:, 0]) > 0 else 0             xmax = max(points[:, 0]) if max(points[:, 0]) > 0 else 0             ymin = min(points[:, 1]) if min(points[:, 1]) > 0 else 0             ymax = max(points[:, 1]) if max(points[:, 1]) > 0 else 0             label = multi["label"]             if xmax <= xmin:                 pass             elif ymax <= ymin:                 pass             else:                 cls_id = classes.index(label)                 b = (float(xmin), float(xmax), float(ymin), float(ymax))                 bb = convert((width, height), b)                 out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')                 print(json_filename, xmin, ymin, xmax, ymax, cls_id) ChangeToYolo5(train_files, "train") ChangeToYolo5(val_files, "val") ChangeToYolo5(test_files, "test")

import shutilimport os file_List = ["train", "val", "test"]for file in file_List:    if not os.path.exists('../VOC/images/%s' % file):        os.makedirs('../VOC/images/%s' % file)    if not os.path.exists('../VOC/labels/%s' % file):        os.makedirs('../VOC/labels/%s' % file)    print(os.path.exists('../tmp/%s.txt' % file))    f = open('../tmp/%s.txt' % file, 'r')    lines = f.readlines()    for line in lines:        print(line)        line = "/".join(line.split('/')[-5:]).strip()        shutil.copy(line, "../VOC/images/%s" % file)        line = line.replace('JPEGImages', 'labels')        line = line.replace('jpg', 'txt')        shutil.copy(line, "../VOC/labels/%s/" % file)

打开voc.yaml文件，修改里面的配置参数

train: VOC/images/train/  # 训练集图片的路径

val: VOC/images/val/  # 验证集图片的路径

# number of classes nc: 2 #检测的类别，本次数据集有两个类别所以写2   # class names names: ["aircraft", "oiltank"]#类别的名称，和转换数据集时的list对应

cfg参数是YoloV5 模型的配置文件，模型的文件存放在models文件夹下面，按照需求填写不同的文件。

weights参数是YoloV5的预训练模型，和cfg对应，例：cfg配置的是yolov5s.yaml，weights就要配置yolov5s.pt

data是配置数据集的配置文件，我们选用的是voc.yaml，所以配置data/voc.yaml

修改上面三个参数就可以开始训练了，其他的参数根据自己的需求修改。修改后的参数配置如下：

parser.add_argument('--weights', type=str, default='yolov5s.pt', help='initial weights path') parser.add_argument('--cfg', type=str, default='yolov5s.yaml', help='model.yaml path') parser.add_argument('--data', type=str, default='data/voc.yaml', help='data.yaml path')

parser = argparse.ArgumentParser(prog='test.py')parser.add_argument('--weights', nargs='+', type=str, default='runs/exp7/weights/best.pt', help='model.pt path(s)')parser.add_argument('--data', type=str, default='data/voc.yaml', help='*.data path')parser.add_argument('--batch-size', type=int, default=2, help='size of each image batch')parser.add_argument('--save-txt', default='True', action='store_true', help='save results to *.txt')