YOLO 图像识别及C++配置

yolo识别速度较快，如何使用yolo进行训练，预测和c++工程使用，本篇文章将详细讲一讲。

一.下载YOLO

从Ultralytics网站上https://github.com/ultralytics/ultralytics 上下载源代码。

当然首先得配置环境，

1.首先下载miniconda3，我下载的是 https://mirrors.tuna.tsinghua.edu.cn/anaconda/miniconda/页面上得Miniconda3-py310_22.11.1-1-Windows-x86_64.exe，安装的时候选择path加到环境变量中。

2.安装cuda和驱动，首先下载驱动，安装驱动，然后下载cuda，cuda版本是11.6，那就下载cuda11.6 https://developer.nvidia.com/cuda-11-6-0-download-archive?target_os=Windows

3.安装pytorch，当然得和cuda版本适配，我下载的是https://pytorch.org/get-started/previous-versions/?_gl=1*6kaf7a*_up*MQ..*_ga*MTgxNzc2OTE1NS4xNzc2MDAxMTMz*_ga_469Y0W5V62*czE3NzYwMDExMzIkbzEkZzAkdDE3NzYwMDExMzIkajYwJGwwJGgw

pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116

4.然后conda 创建一个虚拟环境。

conda create -n myyolov8

二.安装Ultralytics

进入 下载的ultrlytics文件夹中，然后使用

pip install -r requirements.txt

安装所需，最后使用

pip install -e .

这样就安装好了，没遇到问题。

三、使用已有数据集进行训练和使用

3.1训练、使用和导出onnx

from ultralytics import YOLO

if __name__ == '__main__':
    # Create a new YOLO model from scratch
    model = YOLO("yolo26n.yaml")

    # Load a pretrained YOLO model (recommended for training)
    model = YOLO("yolo26n.pt")

    # Train the model using the 'coco8.yaml' dataset for 3 epochs
    results = model.train(data="coco8.yaml", epochs=3)

    # Evaluate the model's performance on the validation set
    results = model.val()

    # Perform object detection on an image using the model
    results = model("https://ultralytics.com/images/bus.jpg")

    # Export the model to ONNX format
    success = model.export(format="onnx")

调用c++

使用vs新建一个工程，配置好opencv

然后利用这几个文件，组建一个工程，

工程中，导入的onnx就是刚生成的onnx，生成的结果就是这个。

四、使用自己的数据集来训练和使用

4.1 目标识别

4.1.1 使用labelme 构建自己的数据集

我选择的是口罩数据集，自己使用labelme来标注，安装labelme的时候，需要解决一些问题，比如

安装Labelme 报错TypeError: 'type' object is not subscriptable

https://blog.csdn.net/weixin_44815456/article/details/145763253

https://blog.csdn.net/weixin_45932490/article/details/148953184

labelme的使用

https://blog.csdn.net/u011119817/article/details/118731850

labelme --flags flags.txt

flags.txt 内容就是：

face

face_mask

mask

导入图像目录，然后就可以画矩形框标注了。

但是label标注后生成的是json文件，需要转成txt文件。转换代码如下：

import os
import shutil
import numpy as np
import json
from glob import glob
import cv2
from sklearn.model_selection import train_test_split

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 *= dw
    w *= dw
    y *= dh
    h *= dh
    return (x, y, w, h)

def change_2_yolo5(files, txt_Name):
    imag_name = []
    for json_file_ in files:
        json_filename = labelme_path + json_file_ + ".json"
        out_file = open('%s/%s.txt' % (labelme_path, json_file_), 'w')
        json_file = json.load(open(json_filename, "r", encoding="utf-8"))
        imag_name.append(json_file_+'.jpg')
        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"].lower()
            if xmax <= xmin or ymax <= ymin:
                continue
            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')
    return imag_name

def image_txt_copy(files, scr_path, dst_img_path, dst_txt_path):
    for file in files:
        img_path = scr_path + file
        shutil.copy(img_path, dst_img_path + file)
        scr_txt_path = scr_path + file.split('.')[0] + '.txt'
        shutil.copy(scr_txt_path, dst_txt_path + file.split('.')[0] + '.txt')

if __name__ == '__main__':
    classes = ['face','face_mask','mask']

    labelme_path = "E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/JPEGImages/"
    files = glob(labelme_path + "*.json")
    files = [i.replace("\\", "/").split("/")[-1].split(".json")[0] for i in files]
    
    trainval_files, test_files = train_test_split(files, test_size=0.1, random_state=55)
    train_files, val_files = train_test_split(trainval_files, test_size=0.1, random_state=55)

    train_name_list = change_2_yolo5(train_files, "train")
    val_name_list = change_2_yolo5(val_files, "val")
    test_name_list = change_2_yolo5(test_files, "test")

    file_List = ["train", "val", "test"]
    for file in file_List:
        if not os.path.exists('E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/images/%s' % file):
            os.makedirs('E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/images/%s' % file)
        if not os.path.exists('E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/labels/%s' % file):
            os.makedirs('E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/labels/%s' % file)

    image_txt_copy(train_name_list, labelme_path, 'E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/images/train/', 'E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/labels/train/')
    image_txt_copy(val_name_list, labelme_path, 'E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/images/val/', 'E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/labels/val/')
    image_txt_copy(test_name_list, labelme_path, 'E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/images/test/', 'E:/BaiduNetdiskDownload/yolov8/yolov8-all-in-one/yolov8-all-in-one/datasets/VOCdevkit/VOC/labels/test/')

然后要制作数据集，使用如下代码，生成 train,val, test

import os, shutil
from sklearn.model_selection import train_test_split

val_size = 0.1
test_size = 0.2
postfix = 'jpg'
imgpath = 'VOCdevkit/JPEGImages'
txtpath = 'VOCdevkit/txt'

os.makedirs('images/train', exist_ok=True)
os.makedirs('images/val', exist_ok=True)
os.makedirs('images/test', exist_ok=True)
os.makedirs('labels/train', exist_ok=True)
os.makedirs('labels/val', exist_ok=True)
os.makedirs('labels/test', exist_ok=True)

listdir = [i for i in os.listdir(txtpath) if 'txt' in i and i != 'classes.txt']
train, test = train_test_split(listdir, test_size=test_size, shuffle=True, random_state=0)
train, val = train_test_split(train, test_size=val_size, shuffle=True, random_state=0)

for i in train:
    shutil.copy('{}/{}.{}'.format(imgpath, i[:-4], postfix), 'images/train/{}.{}'.format(i[:-4], postfix))
    shutil.copy('{}/{}'.format(txtpath, i), 'labels/train/{}'.format(i))

for i in val:
    shutil.copy('{}/{}.{}'.format(imgpath, i[:-4], postfix), 'images/val/{}.{}'.format(i[:-4], postfix))
    shutil.copy('{}/{}'.format(txtpath, i), 'labels/val/{}'.format(i))

for i in test:
    shutil.copy('{}/{}.{}'.format(imgpath, i[:-4], postfix), 'images/test/{}.{}'.format(i[:-4], postfix))
    shutil.copy('{}/{}'.format(txtpath, i), 'labels/test/{}'.format(i))

清除空行

import os

# 设置文件夹路径
folder_path = 'VOCdevkit/txt'

# 获取文件夹中所有txt文件的路径
txt_files = [os.path.join(folder_path, f) for f in os.listdir(folder_path) if f.endswith('.txt')]

# 循环处理每个txt文件
for file_path in txt_files:
    # 读取文件内容
    with open(file_path, 'r') as f:
        lines = f.readlines()

    # 移除空行和只有空白字符的行
    lines = [line for line in lines if line.strip()]

    # 写回到文件中
    with open(file_path, 'w') as f:
        f.writelines(lines)

    print(f"已处理文件：{file_path}")

print("处理完成！")

最后就得到了标注的口罩数据集。

然后放到

face_mask也需要改一下，改完后是这个样子：

这个在这里再放一份

最后要训练了同志们：

训练代码：

import os
# 禁用PyTorch的共享内存加载，直接绕过报错
os.environ["PYTORCH_DISABLE_SHM"] = "1"
from ultralytics import YOLO

if __name__ == '__main__':
    model = YOLO("yolov8n.pt")  #yolov8s-seg.pt
    models = model.train(data="face_mask.yaml",epochs=130,imgsz=640,  batch=1,device =0,workers=0)
    model.export(format="onnx")
    results =models("D:/projection/ultralytics-v8.4.0/datasets/face_mask/images/test/picture_016.jpg")
    results[0].show()

直接就可以训练，当然会遇到内存也不足的问题，还有使用的问题，看这个

错误：The "freeze_support()" line can be omitted if the program is not going to be frozen

https://blog.csdn.net/jackylzh/article/details/157258123

内存也不足的问题：

https://www.jb51.net/article/275647.htm

接着就可以训练了。在这里找到最后一次生成的pt参数文件，

大家也可以看到onnx文件，

4.2 图像分割

数据用的原始数据集，训练是我自己训练

import os
# 禁用PyTorch的共享内存加载，直接绕过报错
os.environ["PYTORCH_DISABLE_SHM"] = "1"
from ultralytics import YOLO

if __name__ == '__main__':
    model = YOLO("yolov8s-seg.pt")  #yolov8s-seg.pt
    models = model.train(data="crack-seg.yaml",epochs=130,imgsz=640, device =0,workers=0)
    # model.export(format="onnx")
    results =models("D:/projection/ultralytics-v8.4.0/datasets/crack-seg/images/test/1818.rf.7e52c1687bfb625e86c54c9d71a66d99.jpg")
    results[0].show()

五、c++调用（转成onnx格式）

把这几个文件搞过来

再vs中，构建一个工程，得到：

指定onnx的路径，运行，得到最后的结果：