使用mmdetection来训练自己的数据集（visdrone）（二）数据集

数据集下载

https://github.com/VisDrone/VisDrone-Dataset

数据集大小：

• trainset (1.44 GB):
• valset (0.07 GB):
• testset-dev (0.28 GB):

VisDrone2019-DET-train.zip
(下载到D:/ultralytics/ultralytics/datasets/VisDrone目录下并解压)
VisDrone2019-DET-val.zip
(下载到D:/ultralytics/ultralytics/datasets/VisDrone目录下并解压)
VisDrone2019-DET-test-dev.zip
(下载到D:/ultralytics/ultralytics/datasets/VisDrone目录下并解压)
VisDrone2019-DET-test-challenge.zip
(下载到D:/ultralytics/ultralytics/datasets/VisDrone目录下并解压)

convert_visdrone2yolo.py (下载到
D:/ultralytics/ultralytics/datasets/VisDrone目录下)
convert_yolo2visdrone.py
注意：VisDrone目录自己创建

visdrone转变代码

import os
from pathlib import Path

# 定义一个函数用于将VisDrone数据集的标注转换为YOLO格式的标注
def visdrone2yolo(dir):
  from PIL import Image
  from tqdm import tqdm

  # 定义一个内部函数，用于将VisDrone的bbox坐标转换为YOLO格式的坐标
  def convert_box(size, box):
      # VisDrone框转换为YOLO xywh框
      dw = 1. / size[0]
      dh = 1. / size[1]
      return (box[0] + box[2] / 2) * dw, (box[1] + box[3] / 2) * dh, box[2] * dw, box[3] * dh

  # 创建标签目录，如果不存在则创建
  (dir / 'labels').mkdir(parents=True, exist_ok=True)
  # 使用tqdm进度条迭代处理annotations文件夹中的所有txt文件
  pbar = tqdm((dir / 'annotations').glob('*.txt'), desc=f'Converting {dir}')
  for f in pbar:
      # 获取对应图片的大小
      img_size = Image.open((dir / 'images' / f.name).with_suffix('.jpg')).size
      lines = []
      with open(f, 'r') as file:  # 读取annotation.txt文件
          for row in [x.split(',') for x in file.read().strip().splitlines()]:
              if row[4] == '0':  # 忽略VisDrone中的‘ignored regions’类别0
                  continue
              cls = int(row[5]) - 1
              box = convert_box(img_size, tuple(map(int, row[:4])))
              lines.append(f"{cls} {' '.join(f'{x:.6f}' for x in box)}\n")
              with open(str(f).replace(f'{os.sep}annotations{os.sep}', f'{os.sep}labels{os.sep}'), 'w') as fl:
                  fl.writelines(lines)  # 将转换后的标签写入label.txt文件

# 指定数据集的根目录
dir = Path("D:/ultralytics/ultralytics/datasets/VisDrone")

# 转换数据集标注
for d in 'VisDrone2019-DET-train', 'VisDrone2019-DET-val', 'VisDrone2019-DET-test-dev':
  visdrone2yolo(dir / d)  # 转换VisDrone标注为YOLO标签

然后通过yolo转coco的代码

import os
import cv2
import json
from tqdm import tqdm
from sklearn.model_selection import train_test_split
import argparse

# 解析命令行参数
parser = argparse.ArgumentParser()
parser.add_argument('--root_dir', default='./dataset/valid', type=str, help="根目录路径，包括./images和./labels及classes.txt")
parser.add_argument('--save_path', type=str, default='./valid.json', help="如果不划分数据集，则提供一个json文件路径")
parser.add_argument('--random_split', action='store_true', help="随机划分数据集，默认比例为8:1:1")
parser.add_argument('--split_by_file', action='store_true', help="根据文件划分数据集，包括./train.txt、./val.txt、./test.txt")

arg = parser.parse_args()

def train_test_val_split_random(img_paths, ratio_train=0.8, ratio_test=0.1, ratio_val=0.1):
    # 确保三个比例之和为1
    assert int(ratio_train + ratio_test + ratio_val) == 1
    train_img, middle_img = train_test_split(img_paths, test_size=1 - ratio_train, random_state=233)
    ratio = ratio_val / (1 - ratio_train)
    val_img, test_img = train_test_split(middle_img, test_size=ratio, random_state=233)
    print("训练集:验证集:测试集 = {}:{}:{}".format(len(train_img), len(val_img), len(test_img)))
    return train_img, val_img, test_img

def train_test_val_split_by_files(img_paths, root_dir):
    # 根据train.txt, val.txt, test.txt文件定义训练集、验证集和测试集
    phases = ['train', 'val', 'test']
    img_split = []
    for p in phases:
        define_path = os.path.join(root_dir, f'{p}.txt')
        print(f'从 {define_path} 读取 {p} 数据集定义')
        assert os.path.exists(define_path)
        with open(define_path, 'r') as f:
            img_paths = f.readlines()
            # 取消注释下面一行可以使用绝对路径
            # img_paths = [os.path.split(img_path.strip())[1] for img_path in img_paths]
            img_split.append(img_paths)
    return img_split[0], img_split[1], img_split[2]

def yolo2coco(arg):
    root_path = arg.root_dir
    print("从路径加载数据", root_path)

    assert os.path.exists(root_path)
    originLabelsDir = os.path.join(root_path, 'labels')
    originImagesDir = os.path.join(root_path, 'images')
    with open(os.path.join(root_path, 'classes.txt')) as f:
        classes = f.read().strip().split()

    indexes = os.listdir(originImagesDir)

    if arg.random_split or arg.split_by_file:
        train_dataset = {'categories': [], 'annotations': [], 'images': []}
        val_dataset = {'categories': [], 'annotations': [], 'images': []}
        test_dataset = {'categories': [], 'annotations': [], 'images': []}

        for i, cls in enumerate(classes, 0):
            category_info = {'id': i, 'name': cls, 'supercategory': 'mark'}
            train_dataset['categories'].append(category_info)
            val_dataset['categories'].append(category_info)
            test_dataset['categories'].append(category_info)

        if arg.random_split:
            print("划分模式: 随机划分")
            train_img, val_img, test_img = train_test_val_split_random(indexes, 0.8, 0.1, 0.1)
        elif arg.split_by_file:
            print("划分模式: 根据文件划分")
            train_img, val_img, test_img = train_test_val_split_by_files(indexes, root_path)
    else:
        dataset = {'categories': [], 'annotations': [], 'images': []}
        for i, cls in enumerate(classes, 0):
            dataset['categories'].append({'id': i, 'name': cls, 'supercategory': 'mark'})

    ann_id_cnt = 0
    for k, index in enumerate(tqdm(indexes)):
        txtFile = index.replace('images', 'txt').replace('.jpg', '.txt').replace('.png', '.txt')
        im = cv2.imread(os.path.join(root_path, 'images/') + index)
        height, width, _ = im.shape
        if arg.random_split or arg.split_by_file:
            if index in train_img:
                dataset = train_dataset
            elif index in val_img:
                dataset = val_dataset
            elif index in test_img:
                dataset = test_dataset
        dataset['images'].append({'file_name': index, 'id': k, 'width': width, 'height': height})
        if not os.path.exists(os.path.join(originLabelsDir, txtFile)):
            continue
        with open(os.path.join(originLabelsDir, txtFile), 'r') as fr:
            labelList = fr.readlines()
            for label in labelList:
                label = label.strip().split()
                x = float(label[1])
                y = float(label[2])
                w = float(label[3])
                h = float(label[4])
                H, W, _ = im.shape
                x1 = (x - w / 2) * W
                y1 = (y - h / 2) * H
                x2 = (x + w / 2) * W
                y2 = (y + h / 2) * H
                cls_id = int(label[0])
                width = max(0, x2 - x1)
                height = max(0, y2 - y1)
                dataset['annotations'].append({
                    'area': width * height,
                    'bbox': [x1, y1, width, height],
                    'category_id': cls_id,
                    'id': ann_id_cnt,
                    'image_id': k,
                    'iscrowd': 0,
                    'segmentation': [[x1, y1, x2, y1, x2, y2, x1, y2]]
                })
                ann_id_cnt += 1

    folder = os.path.join(root_path, 'annotations')
    if not os.path.exists(folder):
        os.makedirs(folder)
    if arg.random_split or arg.split_by_file:
        for phase in ['train', 'val', 'test']:
            json_name = os.path.join(root_path, 'annotations/{}.json'.format(phase))
            with open(json_name, 'w') as f:
                if phase == 'train':
                    json.dump(train_dataset, f)
                elif phase == 'val':
                    json.dump(val_dataset, f)
                elif phase == 'test':
                    json.dump(test_dataset, f)
            print('已保存注释到 {}'.format(json_name))
    else:
        json_name = os.path.join(root_path, 'annotations/{}'.format(arg.save_path))
        with open(json_name, 'w') as f:
            json.dump(dataset, f)
            print('已保存注释到 {}'.format(json_name))

if __name__ == "__main__":
    yolo2coco(arg)

运行代码

python yolo2coco.py --root_dir VisDrone2019-DET-train --save_path train.json
python yolo2coco.py --root_dir VisDrone2019-DET-val --save_path val.json
python yolo2coco.py --root_dir VisDrone2019-DET-test-dev --save_path test.json