yolov8训练自己的实例分割数据集

参考：
1、官方教程：https://docs.ultralytics.com/zh/tasks/segment/

2、YOLOv8制作自己的实例分割数据集保姆级教程（包含json转txt）

下载最新yolov8代码

git clone https://github.com/ultralytics/ultralytics.git

一、制作自己的数据集

省略

二、分离img和json 以及json转yolo格式

2.1 分离images和json文件

使用labelme标注完成后的结果是在一个文件夹中。

分离开来
fenli_image_and_label.py

# 分离jpg txt和json文件


import os.path
import xml.etree.ElementTree as ET
from xml.dom.minidom import Document
import os
import os.path
from PIL import Image

import os
import shutil

class_name = './class.txt'

class_new_name = './class_new.txt'
Class_Name = []
Class_new_Name = []
# img_path  = r'/media/huifeimao/2/diyun/11_h264_data/苏州申通5/images/'    #图片文件夹
# xml_path  = r'/media/huifeimao/2/diyun/11_h264_data/苏州申通5/xml/'  #xml文件存放地址
# ann_path  = r'/media/huifeimao/2/diyun/11_h264_data/苏州申通5/0/'# yolov3标注.txt文件夹

source_folder="/media/diyun/T9/diyun/10_train_data/9_Chinese_chess/key_point/24_0125_train_183/images_src/"

dst_folder="/media/diyun/T9/diyun/10_train_data/9_Chinese_chess/key_point/24_0125_train_183"


jpg_destination_folder=dst_folder + "/images"  # 源文件夹路径
txt_destination_folder = dst_folder + "/labels"  # 目标文件夹路径
json_destination_folder= dst_folder + "/json"  # 目标文件夹路径

is_delete_src=0



def copy_txt_files(source_folder, destination_folder,is_delete_src):
    # 遍历源文件夹中的所有文件和目录
    for root, dirs, files in os.walk(source_folder):
        for file in files:
            # 检查文件扩展名是否为 .txt
            if file.endswith(".txt"):
                # 构建源文件的完整路径
                source_path = os.path.join(root, file)
                # 构建目标文件的完整路径
                destination_path = os.path.join(destination_folder, file)
                # 复制文件到目标文件夹
                shutil.copy(source_path, destination_path)
                if is_delete_src:
                    os.remove(source_path)

def copy_json_files(source_folder, destination_folder,is_delete_src):
    # 遍历源文件夹中的所有文件和目录
    for root, dirs, files in os.walk(source_folder):
        for file in files:
            # 检查文件扩展名是否为 .txt
            if file.endswith(".json"):
                # 构建源文件的完整路径
                source_path = os.path.join(root, file)
                # 构建目标文件的完整路径
                destination_path = os.path.join(destination_folder, file)
                # 复制文件到目标文件夹
                shutil.copy(source_path, destination_path)
                if is_delete_src:
                    os.remove(source_path)

def copy_jpg_files(source_folder, destination_folder):
    # 遍历源文件夹中的所有文件和目录
    for root, dirs, files in os.walk(source_folder):
        for file in files:
            # 检查文件扩展名是否为 .txt
            if file.endswith(".jpg"):
                # 构建源文件的完整路径
                source_path = os.path.join(root, file)
                # 构建目标文件的完整路径
                destination_path = os.path.join(destination_folder, file)
                # 复制文件到目标文件夹
                shutil.copy(source_path, destination_path)




if not os.path.exists(txt_destination_folder):
    os.mkdir(txt_destination_folder)
if not os.path.exists(json_destination_folder):
    os.mkdir(json_destination_folder)
if not os.path.exists(jpg_destination_folder):
    os.mkdir(jpg_destination_folder)


copy_txt_files(source_folder, txt_destination_folder,is_delete_src)
copy_json_files(source_folder, json_destination_folder,is_delete_src)

copy_jpg_files(source_folder, jpg_destination_folder)

2.2、 json转yolo格式

新建segment_json_2_txt.py文件，将代码中的文件路径修改为自己的路径。

❗❗❗代码中第43行的classes中存放的是自己数据集的分类标签，记得修改成自己的。

'''
yolov8训练自己的实例分割数据集
 json转yolo格式
'''


import json
import os
from tqdm import tqdm
 
 
def convert_label(json_dir, save_dir, classes):
    json_paths = os.listdir(json_dir)
    classes = classes.split(',')
 
    for json_path in tqdm(json_paths):
 
        path = os.path.join(json_dir, json_path)
        with open(path, 'r') as load_f:
            json_dict = json.load(load_f)
        h, w = json_dict['imageHeight'], json_dict['imageWidth']
 
        # save txt path
        txt_path = os.path.join(save_dir, json_path.replace('json', 'txt'))
        txt_file = open(txt_path, 'w')
 
        for shape_dict in json_dict['shapes']:
            label = shape_dict['label']
            label_index = classes.index(label)
            points = shape_dict['points']
 
            points_nor_list = []
 
            for point in points:
                points_nor_list.append(point[0] / w)
                points_nor_list.append(point[1] / h)
 
            points_nor_list = list(map(lambda x: str(x), points_nor_list))
            points_nor_str = ' '.join(points_nor_list)
 
            label_str = str(label_index) + ' ' + points_nor_str + '\n'
            txt_file.writelines(label_str)
 
 
if __name__ == "__main__":
 
    json_dir = '/media/diyun/T9/diyun/10_train_data/9_Chinese_chess/segment/24_0302_train_333/json'
    save_dir = '/media/diyun/T9/diyun/10_train_data/9_Chinese_chess/segment/24_0302_train_333/labels'
    classes = 'checkerboard'
 
    convert_label(json_dir, save_dir, classes)

2.3、划分训练和验证集和测试集

将segment_make_train_val_test_txt.py代码中的文件路径修改为自己的路径。

'''
yolov8训练自己的实例分割数据集
yolov8训练，数据集的准备，从yolo的txt 文件，分为预测训练验证
'''


import os
import random


train_percent = 0.9
trainval_percent = 0.1


label_filepath = '/media/diyun/T9/diyun/10_train_data/9_Chinese_chess/segment/24_0302_train_333/labels'  #xml文件存放地址
Imgpath = '/media/diyun/T9/diyun/10_train_data/9_Chinese_chess/segment/24_0302_train_333/images'    #图片文件夹



if not os.path.exists('ImageSets/'):
    os.makedirs('ImageSets/')

total_xml = os.listdir(label_filepath)
num = len(total_xml)
list = range(num)
tv = int(num * trainval_percent)
tr = int(tv * train_percent)
print("trainval_percent=",tv)
print("train_percent=",tr)
trainval = random.sample(list, tv)
train = random.sample(trainval, tr)
ftrainval = open('ImageSets/trainval.txt', 'w')
ftest = open('ImageSets/test.txt', 'w')
ftrain = open('ImageSets/train.txt', 'w')
fval = open('ImageSets/val.txt', 'w')


for i in list:
    name = total_xml[i][:-4] + '\n'
    if i in trainval:
        ftrainval.write(name)
        if i in train:
            ftest.write(name)
        else:
            fval.write(name)
    else:
        ftrain.write(name)
ftrainval.close()
ftrain.close()
fval.close()
ftest.close()


sets = ['train', 'test','val']


ImageSets_path='ImageSets/'

print(Imgpath)
for image_set in sets:
    if not os.path.exists('labels/'):
        os.makedirs('labels/')
    image_ids = open(ImageSets_path+'%s.txt' % (image_set)).read().strip().split()
    list_file = open('%s.txt' % (image_set), 'w')
    for image_id in image_ids:
        list_file.write(Imgpath+'/%s.jpg\n' % (image_id))
    list_file.close()

生成如下：

扫描二维码关注公众号，回复： 17530882 查看本文章

2.4、下载预训练模型

https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8s-seg.pt

https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-seg.pt

三、训练

训练配置文件

mytrain_segment_chess_board.yaml

train: /home/diyun/work/python_project/23_0130_xiangqi_yolov5/yolov8/datasets/train.txt  # 生成的train.txt的路径
val: /home/diyun/work/python_project/23_0130_xiangqi_yolov5/yolov8/datasets/val.txt   # 生成的val.txt的路径


# Classes
names:
  0: chess_board

训练代码

yolov8_segment_train.py

from ultralytics import YOLO
 
# 加载模型
#model = YOLO('yolov8s-pose.yaml')  # build a new model from YAML
model = YOLO('./weights/yolov8s-seg.pt')  # load a pretrained model (recommended for training)
#model = YOLO('yolov8s-pose.yaml').load('yolov8s-pose.pt')  # build from YAML and transfer weights
 
# 训练
results = model.train(data='./datasets/mytrain_segment_chess_board.yaml', 
                      epochs=100, imgsz=640, workers=0, batch=8, project="Chinese_chess", name="segment")

运行

python yolov8_segment_train.py

或者直接命令行

yolo pose train data=./datasets/mytrain_segment_chess_board.yaml model=yolov8s-seg.pt epochs=100 imgsz=640