deerwork-ai
精选
deer-workflow
An open-source Dynamic Workflow runtime that combines deterministic TypeScript orchestration with replaceable Agent runtimes.
AI 仓库情报摘要
FR-AI / ANALYSIS为什么值得关注
该系统将疲劳检测、分心识别、吸烟和手机使用检测与本地部署的大语言模型相结合，实现了语音指令和对话式AI助手，且完全开源。
适合谁使用
- 开发智能座舱或驾驶辅助系统的开发者
- 从事驾驶员行为分析和人机交互的研究人员
- 汽车爱好者和DIY改装玩家
- 需要可定制监控原型的车队安全管理人员
典型使用场景
- 个人驾驶助手，提供疲劳警报和安全评分
- 车队驾驶员风险分析与违规记录的原型系统
- 演示视觉与LLM集成的汽车教育工具
- 测试本地大语言模型在实时语音控制应用中的可行性
项目优势
- 通过MediaPipe和YOLOv8实现对疲劳、分心、吸烟和手机使用的全面驾驶员状态检测
- 本地部署Qwen-7B（通过Ollama）保障隐私和离线语音交互
- 基于LBPH的人脸认证及备选密码实现访问控制
- AI副驾驶自动生成行程总结和安全评分回顾
使用前须知
- 需要摄像头、麦克风以及运行Qwen-7B模型的本地Ollama服务，资源消耗较高
- 语音命令和界面主要为中文，限制了语言可访问性
- 依赖特定YOLOv8权重文件（yolov8_smoking.pt, yolov8n.pt），仓库未直接提供，需自行获取
README 快速开始
🚗 Intelligent Voice Driving System An intelligent in-vehicle driving assistance and real-time monitoring system that integrates computer vision, multimodal behavior monitoring, voice-interactive control, and a local large language model (Qwen-7B).
✨ Features 🎙️ Pure Voice-Interactive Control
Background microphone listening allows voice commands to control camera states, toggle safety monitoring, and query safety scores.
Powered by a local Qwen-7B large language model (via Ollama), it handles system commands and acts as an intelligent AI co-pilot for casual conversation and relief from driving fatigue.
👀 Comprehensive Driver State Monitoring (MediaPipe + YOLOv8)
Fatigue Monitoring: Calculates Eye Aspect Ratio (EAR) to track drowsiness and Mouth Aspect Ratio (MAR) to capture yawning.
Distraction Detection: Utilizes gaze ratio tracking and head pose yaw to detect looking away or wandering attention.
Dangerous Habit Recognition: Combines YOLOv8 models to precisely detect smoking and mobile phone usage while driving.
Head Posture Detection: Monitors hazardous driving postures such as lowering the head due to fatigue.
🤖 Automated Driving Handover Simulation
Triggers emergency alerts and simulates automated driving intervention when extreme fatigue (such as prolonged eye closure or heavy head dropping) exceeds set thresholds.
👤 Facial Authentication & Access Control
Built-in LBPH face recognition automatically verifies the owner's identity upon starting safety monitoring, with password authorization fallback for unrecognized users.
📊 Safety Scoring & Intelligent Trip Summary
Real-time score deduction system (out of 100 points) with timestamped violation logs. Upon trip completion, the AI co-pilot automatically generates a detailed review and professional evaluation of the driving session.
🛠️ Tech Stack GUI Framework: PySide6 (Qt for Python)
Computer Vision & Tracking: OpenCV, MediaPipe (Holistic face, hand, and pose landmarks)
Object Detection: Ultralytics YOLOv8 (for smoking and phone detection)
Voice Technologies: SpeechRecognition (ASR), pyttsx3 (Offline TTS)
Large Language Model: Ollama + Qwen-7B (Local deployment ensuring privacy and offline capability)
📂 Project Structure Plaintext Intelligent-Voice-Driving-System/ │ ├── a1.py # Main program entry (UI logic, vision threads, ASR and LLM scheduling) ├── testui.py
项目描述
This is a real-time driver state monitoring system with pure voice control for smart vehicles, equipped with the Qwen-7B large language model, capable of AI conversation and fully controlled by AI throughout the entire process.
相关仓库与替代方案
根据分类、Topic 和编程语言匹配的相似项目。
slvDev
精选
esp32-ai
A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.
gavamedia
精选
deltafin
Deltafin is a research project that runs the 2.8-trillion-parameter Mixture-of-Experts model Kimi K3 on a single Apple Silicon Mac (e.g., M1 Max with 64 GB) at about 16 seconds per token, using exact, reproducible inference with local or streaming expert loading.