如何构建基于Yolov8Dota数据集的检测系统-尧图网络科技

如何构建基于Yolov8Dota数据集的检测系统
无人机小目标检测适用于YOLO格式的无人机小目标数据集。
2.类别：飞机、轮船、篮球场、大型车辆、小型车辆等15类。
3.数量：5297张

Dota数据集已转化为yolo格式可直接用yolo开始训练，有15个类别，训练15749 验证5297

如果你这孩子已经将 Dota 数据集转化为 YOLO 格式，并且有 15 个类别，训练集包含 15749 张图像，验证集包含 5297 张图像，那么我们可以直接开始使用 YOLOv8 进行训练。以下是详细的步骤和代码示例。

声明：文章内所有代码仅供参考！

项目概述

我们将使用 YOLOv8 训练一个 Dota 目标检测模型。以下是详细的步骤：

数据集准备：确保数据集格式正确。
环境部署：安装必要的库。
模型训练：使用 YOLOv8 训练模型。
指标可视化：查看训练过程中的各项指标。
PyQt5 界面设计：创建一个简单的 GUI 应用来进行预测。

数据集结构

假设你的数据集已经准备好，并且是以 YOLO 格式存储的。以下是数据集的标准结构：

dataset/ ├── images/ │ ├── train/ │ │ ├── image1.jpg │ │ ├── image2.jpg │ │ └── ... │ ├── val/ │ │ ├── image3.jpg │ │ ├── image4.jpg │ │ └── ... │ └── test/ │ ├── image5.jpg │ ├── image6.jpg │ └── ... ├── labels/ │ ├── train/ │ │ ├── image1.txt │ │ ├── image2.txt │ │ └── ... │ ├── val/ │ │ ├── image3.txt │ │ ├── image4.txt │ │ └── ... │ └── test/ │ ├── image5.txt │ ├── image6.txt │ └── ... └── dataset.yaml

dataset.yaml内容如下：

train:./images/trainval:./images/valnc:15names:['hero1','hero2','hero3','hero4','hero5','hero6','hero7','hero8','hero9','hero10','hero11','hero12','hero13','hero14','hero15']

每个图像对应的标签文件是一个文本文件，每行表示一个边界框，格式为：

<class_id> <x_center> <y_center> <width> <height>

环境部署说明

首先，确保你已经安装了必要的库。以下是详细的环境部署步骤：

安装依赖

# 创建虚拟环境（可选）conda create-ndota_detection_envpython=3.9conda activate dota_detection_env# 安装PyTorchpipinstalltorch==1.9torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu111# 安装其他依赖pipinstallopencv-python pyqt5 ultralytics scikit-learn pandas matplotlib seaborn

模型训练权重和指标可视化展示

我们将使用 YOLOv8 进行训练，并在训练过程中记录各种指标，如 F1 曲线、准确率、召回率、损失曲线和混淆矩阵。

训练脚本`train_yolov8.py`

[<title="Training YOLOv8 on Dota Detection Dataset">]fromultralyticsimportYOLOimportos# Define pathsdataset_path='path/to/dataset'weights_path='runs/train/exp/weights/best.pt'# Create dataset.yamlyaml_content=f""" train:{os.path.join(dataset_path,'images/train')}val:{os.path.join(dataset_path,'images/val')}nc: 15 names: ['hero1', 'hero2', 'hero3', 'hero4', 'hero5', 'hero6', 'hero7', 'hero8', 'hero9', 'hero10', 'hero11', 'hero12', 'hero13', 'hero14', 'hero15'] """withopen(os.path.join(dataset_path,'dataset.yaml'),'w')asf:f.write(yaml_content)# Train YOLOv8model=YOLO('yolov8n.pt')# Load a pretrained model (recommended for training)results=model.train(data=os.path.join(dataset_path,'dataset.yaml'),epochs=100,imgsz=640,save=True)# Save the best weightsbest_weights_path=Path('runs/train/exp/weights/best.pt')shutil.copy(best_weights_path,weights_path)

请将path/to/dataset替换为实际的数据集路径。

指标可视化展示

我们将编写代码来可视化训练过程中的各项指标，包括 F1 曲线、准确率、召回率、损失曲线和混淆矩阵。

可视化脚本`visualize_metrics.py`

[<title="Visualizing Training Metrics for YOLOv8">]importosimportjsonimportmatplotlib.pyplotaspltimportseabornassnsimportnumpyasnpfromsklearn.metricsimportconfusion_matrix,ConfusionMatrixDisplay# Load metricsresults_dir='runs/train/exp'metrics_path=os.path.join(results_dir,'results.json')withopen(metrics_path,'r')asf:results=json.load(f)# Extract metricsloss=[entry['loss']forentryinresultsif'loss'inentry]precision=[entry['metrics/precision(m)']forentryinresultsif'metrics/precision(m)'inentry]recall=[entry['metrics/recall(m)']forentryinresultsif'metrics/recall(m)'inentry]mAP_05=[entry['metrics/mAP50(m)']forentryinresultsif'metrics/mAP50(m)'inentry]# Plot loss curveplt.figure(figsize=(15,5))plt.subplot(1,3,1)plt.plot(loss,label='Loss')plt.xlabel('Epochs')plt.ylabel('Loss')plt.title('Training Loss Curve')plt.legend()# Plot precision and recall curvesplt.subplot(1,3,2)plt.plot(precision,label='Precision')plt.plot(recall,label='Recall')plt.xlabel('Epochs')plt.ylabel('Score')plt.title('Precision and Recall Curves')plt.legend()# Plot mAP@0.5 curveplt.subplot(1,3,3)plt.plot(mAP_05,label='mAP@0.5')plt.xlabel('Epochs')plt.ylabel('mAP@0.5')plt.title('mAP@0.5 Curve')plt.legend()plt.tight_layout()plt.show()# Confusion matrix# Assuming you have predictions and true labels# For demonstration, let's create some dummy datanum_classes=15true_labels=np.random.randint(0,num_classes,size=100)# Random true labelspredictions=np.random.randint(0,num_classes,size=100)# Random predicted labelscm=confusion_matrix(true_labels,predictions,labels=list(range(num_classes)))labels=[f'Hero{i+1}'foriinrange(num_classes)]disp=ConfusionMatrixDisplay(confusion_matrix=cm,display_labels=labels)disp.plot(cmap=plt.cm.Blues)plt.title('Confusion Matrix')plt.xticks(rotation=90)plt.yticks(rotation=0)plt.tight_layout()plt.show()

PyQt5设计的界面

我们将使用 PyQt5 设计一个简单的 GUI 界面来进行模型预测。

GUI代码`gui_app.py`

[<title="PyQt5 GUI for YOLOv8 Dota Detection">]importsysimportcv2importnumpyasnpfromPyQt5.QtWidgetsimportQApplication,QMainWindow,QLabel,QPushButton,QVBoxLayout,QWidget,QFileDialog,QMessageBox,QProgressBar,QTextEditfromPyQt5.QtGuiimportQImage,QPixmapfromPyQt5.QtCoreimportQTimerfromultralyticsimportYOLOclassMainWindow(QMainWindow):def__init__(self):super().__init__()self.setWindowTitle("Dota Hero Detection")self.setGeometry(100,100,800,600)self.central_widget=QWidget(self)self.setCentralWidget(self.central_widget)self.layout=QVBoxLayout(self.central_widget)self.label_display=QLabel(self)self.layout.addWidget(self.label_display)self.button_layout=QHBoxLayout()self.pushButton_image=QPushButton("Open Image",self)self.pushButton_image.clicked.connect(self.open_image)self.button_layout.addWidget(self.pushButton_image)self.pushButton_folder=QPushButton("Open Folder",self)self.pushButton_folder.clicked.connect(self.open_folder)self.button_layout.addWidget(self.pushButton_folder)self.pushButton_video=QPushButton("Open Video",self)self.pushButton_video.clicked.connect(self.open_video)self.button_layout.addWidget(self.pushButton_video)self.pushButton_camera=QPushButton("Start Camera",self)self.pushButton_camera.clicked.connect(self.start_camera)self.button_layout.addWidget(self.pushButton_camera)self.pushButton_stop=QPushButton("Stop Camera",self)self.pushButton_stop.clicked.connect(self.stop_camera)self.button_layout.addWidget(self.pushButton_stop)self.layout.addLayout(self.button_layout)self.model=YOLO('runs/train/exp/weights/best.pt')self.cap=Noneself.timer=QTimer()self.timer.timeout.connect(self.process_frame)defload_image(self,file_name):img=cv2.imread(file_name)# BGRassertimgisnotNone,f'Image Not Found{file_name}'returnimgdefprocess_image(self,img):results=self.model(img,stream=True)forresultinresults:boxes=result.boxes.cpu().numpy()forboxinboxes:r=box.xyxy[0].astype(int)cls=int(box.cls[0])conf=box.conf[0]label=f'{self.model.names[cls]}{conf:.2f}'color=(0,255,0)# Greencv2.rectangle(img,r[:2],r[2:],color,2)cv2.putText(img,label,(r[0],r[1]-10),cv2.FONT_HERSHEY_SIMPLEX,0.9,color,2)rgb_image=cv2.cvtColor(img,cv2.COLOR_BGR2RGB)h,w,ch=rgb_image.shape bytes_per_line=ch*w qt_image=QImage(rgb_image.data,w,h,bytes_per_line,QImage.Format_RGB888)pixmap=QPixmap.fromImage(qt_image)self.label_display.setPixmap(pixmap.scaled(800,600))defopen_image(self):options=QFileDialog.Options()file_name,_=QFileDialog.getOpenFileName(self,"QFileDialog.getOpenFileName()","","Images (*.jpeg *.jpg);;All Files (*)",options=options)iffile_name:img=self.load_image(file_name)self.process_image(img)defopen_folder(self):folder_name=QFileDialog.getExistingDirectory(self,"Select Folder")iffolder_name:forfilenameinos.listdir(folder_name):iffilename.lower().endswith(('.png','.jpg','.jpeg')):file_path=os.path.join(folder_name,filename)img=self.load_image(file_path)self.process_image(img)defopen_video(self):options=QFileDialog.Options()file_name,_=QFileDialog.getOpenFileName(self,"QFileDialog.getOpenFileName()","","Videos (*.mp4 *.avi);;All Files (*)",options=options)iffile_name:self.cap=cv2.VideoCapture(file_name)self.timer.start(30)# Process frame every 30 msdefstart_camera(self):self.cap=cv2.VideoCapture(0)self.timer.start(30)# Process frame every 30 msdefstop_camera(self):ifself.capisnotNone:self.cap.release()self.cap=Noneself.timer.stop()defprocess_frame(self):ifself.capisnotNone:ret,frame=self.cap.read()ifret:self.process_image(frame)else:self.cap.release()self.cap=Noneself.timer.stop()if__name__=="__main__":app=QApplication(sys.argv)window=MainWindow()window.show()sys.exit(app.exec_())

辅助工具文件`utils.py`

这个文件可以用来存放一些辅助函数，比如保存结果等。

[<title="Utility Functions for Dota Hero Detection">]importcv2importosdefsave_results(image,detections,output_dir,filename):fordetindetections:r=det['bbox']cls=det['class']conf=det['confidence']label=f'{cls}{conf:.2f}'color=(0,255,0)# Greencv2.rectangle(image,(int(r[0]),int(r[1])),(int(r[2]),int(r[3])),color,2)cv2.putText(image,label,(int(r[0]),int(r[1])-10),cv2.FONT_HERSHEY_SIMPLEX,0.9,color,2)output_path=os.path.join(output_dir,filename)cv2.imwrite(output_path,image)

运行效果展示

假设你已经有了运行效果的图像，可以在README.md中添加这些图像以供参考。

# Dota Hero Detection System ## Overview This project provides a deep learning-based system for detecting Dota heroes in images using thermal infrared images. The system can identify Dota heroes in images, folders, videos, and live camera feeds. ## Environment Setup - Software: PyCharm + Anaconda - Environment: Python=3.9, OpenCV-Python, PyQt5, Torch=1.9 ## Features - Detects Dota heroes. - Supports detection on images, folders, videos, and live camera feed. - Batch processing of images. - Real-time display of detected heroes with confidence scores and bounding boxes. - Saving detection results. ## Usage 1. Run the program. 2. Choose an option to detect Dota heroes in images, folders, videos, or via the camera. ## Screenshots ![Example Screenshot](data/screenshots/example_screenshot.png)