基于深度学习的数字识别检测系统（YOLOv10+YOLO数据集+UI界面+模型）

train: .\datasets\images\train val: .\datasets\images\val test: .\datasets\images\test nc: 10 names: ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']

数据集制作流程

标注数据：使用标注工具（如LabelImg、CVAT等）对图像中的目标进行标注。每个目标需要标出边界框，并且标注类别。
转换格式：将标注的数据转换为YOLO格式。YOLO标注格式为每行：<object-class> <x_center> <y_center> <width> <height>，这些坐标是相对于图像尺寸的比例。
分割数据集：将数据集分为训练集、验证集和测试集，通常的比例是80%训练集、10%验证集和10%测试集。
准备标签文件：为每张图片生成一个对应的标签文件，确保标签文件与图片的命名一致。
调整图像尺寸：根据YOLO网络要求，统一调整所有图像的尺寸（如416x416或608x608）。

四、项目环境配置

创建虚拟环境

首先新建一个Anaconda环境，每个项目用不同的环境，这样项目中所用的依赖包互不干扰。

终端输入

conda create -n yolov10 python==3.9

激活虚拟环境

conda activate yolov10

安装cpu版本pytorch

pip install torch torchvision torchaudio

from ultralytics import YOLOv10 model_path = 'yolov10s.pt' data_path = 'datasets/data.yaml' if __name__ == '__main__': model = YOLOv10(model_path) results = model.train(data=data_path, epochs=500, batch=64, device='0', workers=0, project='runs/detect', name='exp', )

根据实际情况更换模型 yolov10n.yaml (nano)：轻量化模型，适合嵌入式设备，速度快但精度略低。 yolov10s.yaml (small)：小模型，适合实时任务。 yolov10m.yaml (medium)：中等大小模型，兼顾速度和精度。 yolov10b.yaml (base)：基本版模型，适合大部分应用场景。 yolov10l.yaml (large)：大型模型，适合对精度要求高的任务。

--batch 64：每批次64张图像。
--epochs 500：训练500轮。
--datasets/data.yaml：数据集配置文件。
--weights yolov10s.pt：初始化模型权重，yolov10s.pt是预训练的轻量级YOLO模型。

训练结果

六、核心代码

import sys import cv2 import numpy as np from PyQt5.QtWidgets import QApplication, QMessageBox, QFileDialog from PyQt5.QtCore import QThread, pyqtSignal from ultralytics import YOLOv10 from UiMain import UiMainWindow import time import os class DetectionThread(QThread): frame_received = pyqtSignal(np.ndarray, np.ndarray, list) # 原始帧, 检测帧, 检测结果 finished_signal = pyqtSignal() # 线程完成信号 def __init__(self, model, source, conf, iou, parent=None): super().__init__(parent) self.model = model self.source = source self.conf = conf self.iou = iou self.running = True def run(self): try: if isinstance(self.source, int) or self.source.endswith(('.mp4', '.avi', '.mov')): # 视频或摄像头 cap = cv2.VideoCapture(self.source) while self.running and cap.isOpened(): ret, frame = cap.read() if not ret: break # 保存原始帧 original_frame = frame.copy() # 检测 results = self.model(frame, conf=self.conf, iou=self.iou) annotated_frame = results[0].plot() # 提取检测结果 detections = [] for result in results: for box in result.boxes: class_id = int(box.cls) class_name = self.model.names[class_id] confidence = float(box.conf) x, y, w, h = box.xywh[0].tolist() detections.append((class_name, confidence, x, y)) # 发送信号 self.frame_received.emit( cv2.cvtColor(original_frame, cv2.COLOR_BGR2RGB), cv2.cvtColor(annotated_frame, cv2.COLOR_BGR2RGB), detections ) # 控制帧率 time.sleep(0.03) # 约30fps cap.release() else: # 图片 frame = cv2.imread(self.source) if frame is not None: original_frame = frame.copy() results = self.model(frame, conf=self.conf, iou=self.iou) annotated_frame = results[0].plot() # 提取检测结果 detections = [] for result in results: for box in result.boxes: class_id = int(box.cls) class_name = self.model.names[class_id] confidence = float(box.conf) x, y, w, h = box.xywh[0].tolist() detections.append((class_name, confidence, x, y)) self.frame_received.emit( cv2.cvtColor(original_frame, cv2.COLOR_BGR2RGB), cv2.cvtColor(annotated_frame, cv2.COLOR_BGR2RGB), detections ) except Exception as e: print(f"Detection error: {e}") finally: self.finished_signal.emit() def stop(self): self.running = False class MainWindow(UiMainWindow): def __init__(self): super().__init__() # 初始化模型 self.model = None self.detection_thread = None self.current_image = None self.current_result = None self.video_writer = None self.is_camera_running = False self.is_video_running = False self.last_detection_result = None # 新增：保存最后一次检测结果 # 连接按钮信号 self.image_btn.clicked.connect(self.detect_image) self.video_btn.clicked.connect(self.detect_video) self.camera_btn.clicked.connect(self.detect_camera) self.stop_btn.clicked.connect(self.stop_detection) self.save_btn.clicked.connect(self.save_result) # 初始化模型 self.load_model() def load_model(self): try: model_name = self.model_combo.currentText() self.model = YOLOv10(f"{model_name}.pt") # 自动下载或加载本地模型 self.update_status(f"模型 {model_name} 加载成功") except Exception as e: QMessageBox.critical(self, "错误", f"模型加载失败: {str(e)}") self.update_status("模型加载失败") def detect_image(self): if self.detection_thread and self.detection_thread.isRunning(): QMessageBox.warning(self, "警告", "请先停止当前检测任务") return file_path, _ = QFileDialog.getOpenFileName( self, "选择图片", "", "图片文件 (*.jpg *.jpeg *.png *.bmp)") if file_path: self.clear_results() self.current_image = cv2.imread(file_path) self.current_image = cv2.cvtColor(self.current_image, cv2.COLOR_BGR2RGB) self.display_image(self.original_image_label, self.current_image) # 创建检测线程 conf = self.confidence_spinbox.value() iou = self.iou_spinbox.value() self.detection_thread = DetectionThread(self.model, file_path, conf, iou) self.detection_thread.frame_received.connect(self.on_frame_received) self.detection_thread.finished_signal.connect(self.on_detection_finished) self.detection_thread.start() self.update_status(f"正在检测图片: {os.path.basename(file_path)}") def detect_video(self): if self.detection_thread and self.detection_thread.isRunning(): QMessageBox.warning(self, "警告", "请先停止当前检测任务") return file_path, _ = QFileDialog.getOpenFileName( self, "选择视频", "", "视频文件 (*.mp4 *.avi *.mov)") if file_path: self.clear_results() self.is_video_running = True # 初始化视频写入器 cap = cv2.VideoCapture(file_path) frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)) frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) fps = cap.get(cv2.CAP_PROP_FPS) cap.release() # 创建保存路径 save_dir = "results" os.makedirs(save_dir, exist_ok=True) timestamp = time.strftime("%Y%m%d_%H%M%S") save_path = os.path.join(save_dir, f"result_{timestamp}.mp4") fourcc = cv2.VideoWriter_fourcc(*'mp4v') self.video_writer = cv2.VideoWriter(save_path, fourcc, fps, (frame_width, frame_height)) # 创建检测线程 conf = self.confidence_spinbox.value() iou = self.iou_spinbox.value() self.detection_thread = DetectionThread(self.model, file_path, conf, iou) self.detection_thread.frame_received.connect(self.on_frame_received) self.detection_thread.finished_signal.connect(self.on_detection_finished) self.detection_thread.start() self.update_status(f"正在检测视频: {os.path.basename(file_path)}") def detect_camera(self): if self.detection_thread and self.detection_thread.isRunning(): QMessageBox.warning(self, "警告", "请先停止当前检测任务") return self.clear_results() self.is_camera_running = True # 创建检测线程 (默认使用摄像头0) conf = self.confidence_spinbox.value() iou = self.iou_spinbox.value() self.detection_thread = DetectionThread(self.model, 0, conf, iou) self.detection_thread.frame_received.connect(self.on_frame_received) self.detection_thread.finished_signal.connect(self.on_detection_finished) self.detection_thread.start() self.update_status("正在从摄像头检测...") def stop_detection(self): if self.detection_thread and self.detection_thread.isRunning(): self.detection_thread.stop() self.detection_thread.quit() self.detection_thread.wait() if self.video_writer: self.video_writer.release() self.video_writer = None self.is_camera_running = False self.is_video_running = False self.update_status("检测已停止") def on_frame_received(self, original_frame, result_frame, detections): # 更新原始图像和结果图像 self.display_image(self.original_image_label, original_frame) self.display_image(self.result_image_label, result_frame) # 保存当前结果帧用于后续保存 self.last_detection_result = result_frame # 新增：保存检测结果 # 更新表格 self.clear_results() for class_name, confidence, x, y in detections: self.add_detection_result(class_name, confidence, x, y) # 保存视频帧 if self.video_writer: self.video_writer.write(cv2.cvtColor(result_frame, cv2.COLOR_RGB2BGR)) def on_detection_finished(self): if self.video_writer: self.video_writer.release() self.video_writer = None self.update_status("视频检测完成，结果已保存") elif self.is_camera_running: self.update_status("摄像头检测已停止") else: self.update_status("图片检测完成") def save_result(self): if not hasattr(self, 'last_detection_result') or self.last_detection_result is None: QMessageBox.warning(self, "警告", "没有可保存的检测结果") return save_dir = "results" os.makedirs(save_dir, exist_ok=True) timestamp = time.strftime("%Y%m%d_%H%M%S") if self.is_camera_running or self.is_video_running: # 保存当前帧为图片 save_path = os.path.join(save_dir, f"snapshot_{timestamp}.jpg") cv2.imwrite(save_path, cv2.cvtColor(self.last_detection_result, cv2.COLOR_RGB2BGR)) self.update_status(f"截图已保存: {save_path}") else: # 保存图片检测结果 save_path = os.path.join(save_dir, f"result_{timestamp}.jpg") cv2.imwrite(save_path, cv2.cvtColor(self.last_detection_result, cv2.COLOR_RGB2BGR)) self.update_status(f"检测结果已保存: {save_path}") def closeEvent(self, event): self.stop_detection() event.accept() if __name__ == "__main__": app = QApplication(sys.argv) # 设置应用程序样式 app.setStyle("Fusion") # 创建并显示主窗口 window = MainWindow() window.show() sys.exit(app.exec_())