我是训练的行人目标检测。

首先下载YOLOFastestv2：https://github.com/dog-qiuqiu/FastestDet

yolofastestv2需要这样的数据集：

.
├── category.names        # .names category label file
├── train                 # train dataset
│   ├── 000001.jpg
│   ├── 000001.txt
│   ├── 000002.jpg
│   ├── 000002.txt
│   ├── 000003.jpg
│   └── 000003.txt
├── train.txt              # train dataset path .txt file
├── val                    # val dataset
│   ├── 000043.jpg
│   ├── 000043.txt
│   ├── 000057.jpg
│   ├── 000057.txt
│   ├── 000070.jpg
│   └── 000070.txt
└── val.txt                # val dataset path .txt file

一般我们的数据集是voc样式的数据集即文件夹中只有这两个文件：

train和val文件夹内的0001.txt为yolo数据集的格式，其内容如下：

0 0.344192634561 0.611 0.416430594901 0.262
1 0.509915014164 0.51 0.974504249292 0.972

首先我们将xml转成这个格式的，在Annotations和JPEGImages的同级路径下创建一个py文件：

# 作者：李富贵
# 公众号：猛男技术控
# 输入：xml的文件夹路径
# 输出：在txt路径下创建对应的txt文件
import xml.etree.ElementTree as ET
import osdef convert(size, box):dw = 1. / (size[0])dh = 1. / (size[1])x = (box[0] + box[1]) / 2.0 - 1y = (box[2] + box[3]) / 2.0 - 1w = box[1] - box[0]h = box[3] - box[2]x = x * dww = w * dwy = y * dhh = h * dhreturn x, y, w, hdef convert_annotation(xml_file, txt_path, filename):in_file = open('{}/{}.xml'.format(xml_file, filename), encoding='UTF-8')out_file = open('{}/{}.txt'.format(txt_path, filename), 'w')tree = ET.parse(in_file)root = tree.getroot()size = root.find('size')w = int(size.find('width').text)h = int(size.find('height').text)for obj in root.iter('object'):# difficult = obj.find('difficult').text# difficult = obj.find('Difficult').textif obj.find('difficult'):difficult = obj.find('difficult').textelse:difficult = 0cls = obj.find('name').textif cls not in classes or int(difficult) == 1:continuecls_id = classes.index(cls)xmlbox = obj.find('bndbox')b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text),float(xmlbox.find('ymax').text))b1, b2, b3, b4 = b# 标注越界修正if b2 > w:b2 = wif b4 > h:b4 = hb = (b1, b2, b3, b4)bb = convert((w, h), b)out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')# 改成自己的类别
classes = ["person"]
# 改成自己xml文件夹和要保存的txt文件夹的路径
xml_path = 'Annotations'
txt_path = 'txt'
if not os.path.exists(txt_path):os.makedirs(txt_path)xml_files = os.listdir(xml_path)
for xml_file in xml_files:print(xml_path + '/' + xml_file)convert_annotation(xml_path, txt_path, xml_file[:-4])

然后得到一个txt文件，里面就是存放所有已转换的txt

第二步就是将文件分成train和test

import os
import shutil
import randomall_txt = os.listdir('txt')
random.shuffle(all_txt)trainf = all_txt[:1316]
vlaf = all_txt[1317:]i = 1
for trainimg in trainf:try:shutil.copy(os.path.join('JPEGImages',trainimg[:-3]+'png'),'train')shutil.copy(os.path.join('txt', trainimg), 'train')print(i)i += 1except:print('no such file ',os.path.join('JPEGImages',trainimg[:-3]+'jpg'))
j = 1for valimg in vlaf:try:shutil.copy(os.path.join('JPEGImages',valimg[:-3]+'png'),'v')shutil.copy(os.path.join('txt', valimg), 'v')print(j)j += 1except:print('no such file ',os.path.join('JPEGImages',valimg))

注意自己创建号train和val文件夹，修改train和val文件的个数。
这样就生成了下面这样的文件

├── train
│   ├── 000001.jpg
│   ├── 000001.txt
│   ├── 000002.jpg
│   ├── 000002.txt
│   ├── 000003.jpg
│   └── 000003.txt
└── val├── 000043.jpg├── 000043.txt├── 000057.jpg├── 000057.txt├── 000070.jpg└── 000070.txt

然后将所有文件读入txt中：

import ostraintxt = open('train.txt','a')
val = open('val.txt','a')
for img in os.listdir('train'):if "png" in img:traintxt.write(os.path.join('INRIAPerson-823/train',img)+'\n')print(img)else:print(img,'error')
for img in os.listdir('val'):if "png" in img:valtxt.write(os.path.join('INRIAPerson-823/val',img)+'\n')print(img)else:print(img,'error')
valtxt.close()
traintxt.close()

然后修改data下的.data文件

对应的路径自己修改

然后就是训练了：
先跑下这个：

python3 genanchors.py --traintxt ./train.txt

然后

python3 train.py --data data/coco.data

下不了的关注我公众号：猛男技术控，回复yolofastest 可以提供代码、数据、以及我训练的权重。