基于Yolov8网络进行目标检测（二）-安装和自定义数据集

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

pip install  ultralytics

# Build a new model from YAML and start training from scratch
 
yolo detect train data=coco128.yaml model=yolov8n.yaml epochs=100 imgsz=640
 

 
# Start training from a pretrained *.pt model
 
yolo detect train data=coco128.yaml model=yolov8n.pt epochs=100 imgsz=640
 

 
# Build a new model from YAML, transfer pretrained weights to it and start training
 
yolo detect train data=coco128.yaml model=yolov8n.yaml pretrained=yolov8n.pt epochs=100 imgsz=640 

from ultralytics import YOLO
 

 
# Load a model
 
model = YOLO('yolov8n.yaml') # build a new model from YAML
 
model = YOLO('yolov8n.pt') # load a pretrained model (recommended for training)
 
model = YOLO('yolov8n.yaml').load('yolov8n.pt') # build from YAML and transfer weights
 

 
# Train the model
 
results = model.train(data='coco128.yaml', epochs=100, imgsz=640) 
 

└─VOCdevkit
 
 └─VOC2012
 
 ├─Annotations
 
 ├─ImageSets
 
 │ ├─Action
 
 │ ├─Layout
 
 │ ├─Main
 
 │ └─Segmentation
 
 ├─JPEGImages
 
 ├─SegmentationClass
 
 └─SegmentationObject

import os
 
import random
 
import argparse
 

 
parser = argparse.ArgumentParser()
 
#xml文件的地址，根据自己的数据进行修改 xml一般存放在Annotations下
 
parser.add_argument('--xml_path', default='VOCdevkit/VOC2012/Annotations', type=str, help='input xml label path')
 
#数据集的划分，地址选择自己数据下的ImageSets/Main
 
parser.add_argument('--txt_path', default='VOCdevkit/VOC2012/ImageSets/Main', type=str, help='output txt label path')
 
opt = parser.parse_args()
 
# Namespace(xml_path='VOCdevkit/VOC2012/Annotations', txt_path='VOCdevkit/VOC2012/ImageSets/Main')
 

 
# 训练+验证集一共所占的比例为0.8，剩下的0.2就是测试集
 
# (train+val)/(train+val+test)=80%
 
trainval_percent = 0.8
 
# (train)/(train+val)=80%
 
# 训练集在训练集和验证集总集合中占的比例
 
train_percent = 0.8
 

 
xmlfilepath = opt.xml_path
 
# VOCdevkit/VOC2012/Annotations
 
txtsavepath = opt.txt_path
 
# VOCdevkit/dataset/ImageSets/Main
 

 
# 获取标注文件数量
 
total_xml = os.listdir(xmlfilepath)
 
# 创建文件目录
 
if not os.path.exists(txtsavepath):
 
    os.makedirs(txtsavepath)
 

 
# 随机打散文件序号，生成trainval和train两个随机数组
 
num = len(total_xml)
 
list_index = range(num)
 
tv = int(num * trainval_percent)
 
tr = int(tv * train_percent)
 
trainval = random.sample(list_index, tv)
 
train = random.sample(trainval, tr)
 

 

 
fileTrainVal = open(txtsavepath + '/trainval.txt', 'w')
 
fileTrain = open(txtsavepath + '/train.txt', 'w')
 
fileVal = open(txtsavepath + '/val.txt', 'w')
 
fileTest = open(txtsavepath + '/test.txt', 'w')
 

 
for i in list_index:
 
 # 获取文件名
 
    name = total_xml[i][:-4] + '\n'
 
 # 根据trainval，train，val，test的顺序依次写入相关文件
 
 if i in trainval:
 
        fileTrainVal.write(name)
 
 if i in train:
 
            fileTrain.write(name)
 
 else:
 
            fileVal.write(name)
 
 else:
 
        fileTest.write(name)
 

 
fileTrainVal.close()
 
fileTrain.close()
 
fileVal.close()
 
fileTest.close()

# -*- coding: utf-8 -*-
 
import xml.etree.ElementTree as ET
 
import os
 

 
sets = ['train', 'val', 'test']
 
classes = ["aeroplane", 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', 'dog',
 
 'horse', 'motorcycle', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor']
 
absPath = os.getcwd()
 

 
def convert(size, box):
 
 '''
 

 
 :param size: 图片size
 
 :param box: 标注框坐标
 
 :return:
 
    VOC->YOLO转换算法
 
        norm_x=(xmin + xmax)/2/width
 
        norm_y=(ymin + ymax)/2/height
 
        norm_w=(xmax - xmin)/width
 
        norm_h=(ymax - ymin)/height
 
    YOLO->VOC转换算法
 
        xmin=width * (norm_x - 0.5 * norm_w)
 
        ymin=height * (norm_y - 0.5 * norm_h)
 
        xmax=width * (norm_x + 0.5 * norm_w)
 
        ymax=height * (norm_y + 0.5 * norm_h)
 
 '''
 
    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
 

 

 
def ConvertAnnotation(image_id):
 
    inputFile = open(absPath + '/VOCdevkit/VOC2012/Annotations/%s.xml' % (image_id), encoding='UTF-8')
 
    outFile = open(absPath + '/VOCdevkit/VOC2012/YOLOLabels/%s.txt' % (image_id), 'w')
 
 '''
 
    VOC2012 标注格式
 
 <annotation>
 
 <folder>VOC2012</folder>
 
 <filename>2008_007069.jpg</filename>
 
 <source>
 
 <database>The VOC2008 Database</database>
 
 <annotation>PASCAL VOC2008</annotation>
 
 <image>flickr</image>
 
 </source>
 
 <size>
 
 <width>500</width>
 
 <height>375</height>
 
 <depth>3</depth>
 
 </size>
 
 <segmented>0</segmented>
 
 <object>
 
 <name>sheep</name>
 
 <pose>Right</pose>
 
 <truncated>0</truncated>
 
 <occluded>0</occluded>
 
 <bndbox>
 
 <xmin>411</xmin>
 
 <ymin>172</ymin>
 
 <xmax>445</xmax>
 
 <ymax>195</ymax>
 
 </bndbox>
 
 <difficult>0</difficult>
 
 </object>
 
 <object></object>
 
 </annotation>
 
 '''
 

 
 '''
 
 Yolo 标注文件格式
 
    labelclass xCenter yCenter width height
 
 每个标签有五个数据，依次代表：
 
 所标注内容的类别，数字与类别一一对应
 
 1、labelclass 标注框类别 labelclass
 
 2、xCenter 归一化后标注框的中心点的x轴
 
 3、yCenter 归一化后标注框的中心点的y轴
 
 4、width 归一化后目标框的宽度
 
 5、height 归一化后目标框的高度 
 
 '''
 
    tree = ET.parse(inputFile)
 
    root = tree.getroot()
 
 # 获取标注图片的大小
 
    size = root.find('size')
 
    width = int(size.find('width').text)
 
    height = int(size.find('height').text)
 
 # 获取标注框信息
 
 for obj in root.iter('object'):
 
        difficult = obj.find('difficult').text
 
 # 获取标注类别名称
 
        cls = obj.find('name').text
 
 if cls not in classes or int(difficult) == 1:
 
 continue
 
 # 将标注类别按照classes列表信息转换为索引ID
 
        clsId = classes.index(cls)
 
 # 获取标注框信息
 
        xmlBox = obj.find('bndbox')
 
        boundry = (float(xmlBox.find('xmin').text), float(xmlBox.find('xmax').text), float(xmlBox.find('ymin').text),
 
 float(xmlBox.find('ymax').text))
 
        xmin, xmax, ymin, ymax = boundry
 
 # 标注越界修正
 
 if xmax > width:
 
            xmax = width
 
 if ymax > height:
 
            ymax = height
 
        box = (xmin, xmax, ymin, ymax)
 
        transBox = convert((width, height), box)
 
        outFile.write(str(clsId) + " " + " ".join([str(a) for a in transBox]) + '\n')
 

 
# 判断标注转换目录是否存在
 
if not os.path.exists(absPath + '/VOCdevkit/VOC2012/YOLOLabels/'):
 
    os.makedirs(absPath + '/VOCdevkit/VOC2012/YOLOLabels/')
 
for imageSet in sets:
 
 # 获取当前文件(train/val/test)的图片ID
 
    imageIds = open(absPath + '/VOCdevkit/VOC2012/ImageSets/Main/%s.txt' % (imageSet)).read().strip().split()
 
    listFile = open(absPath + '/VOCdevkit/VOC2012/%s.txt' % (imageSet), 'w')
 
 for imageId in imageIds:
 
 # 遍历文件名列表，分别将图片文件全路径写入新的文件中
 
        listFile.write(absPath + '/VOCdevkit/VOC2012/JPEGImages/%s.jpg\n' % (imageId))
 
 # 进行文件格式转换
 
 ConvertAnnotation(imageId)
 
    listFile.close()