🧠 PDV Vision - AI Visual Guide for the Visually Impaired

PDV Vision is an AI-powered assistant that uses image recognition and speech to help visually impaired users navigate their surroundings with greater confidence.

🌟 Features

• 📷 Object Detection
  Object detection and smart image captioning via YOLOv5 + GPT-4o

• 🗣️ Speech Feedback
  Real-time TTS speech output in Chinese and English

• 📍 Location Awareness
  Location-aware feedback depending on selected mode

• 🧭 Three Operational Modes
  Three modes to optimize recognition and interaction:

  • Outdoor Mode: Detects streets, obstacles, traffic lights
  • Indoor Mode: Detects restrooms, doors, entrances/exits
  • Normal Mode: General-purpose recognition and narration

🔧 Tech Stack

• Frontend: Basic4Android (B4A)
• Backend: Flask + YOLOv5 + OpenAI GPT-4o
• Hosting: Supports local and cloud deployment (e.g., Render)

🚀 Getting Started

📦 Backend Setup

git clone https://github.com/pandavio/vision-server.git
cd pdv-vision-server
pip install -r requirements.txt
python app.py

📱 Android App Setup

• Open PDVVision.b4a with B4A
• Update Starter.ServerURL with your backend address
• Build and install the APK to your Android phone

🔑 Required API Keys

PDV Vision requires the following two API keys to function properly:

• 🧠 OpenAI API Key: Used to call the GPT-4o multimodal model for image understanding and natural language generation
• 🗺️ Google Maps API Key: Used to retrieve GPS location and address details for enhanced outdoor navigation (used only in outdoor mode)

Before running the system, please ensure:

• B4A Client: Add both API keys in the corresponding sections of your Android client code
• Flask Backend: Set the OpenAI API key as an environment variable or in a configuration file (e.g., OPENAI_API_KEY)

📂 Project Structure

PDVVision/
├── B4A/              # Android app (B4A)
├── server/           # Backend server (Flask)
├── README.md
├── requirements.txt

📜 License

This project is licensed under the MIT License.

🧠 PDV Vision - AI视觉导盲助手

PDV Vision 是一款结合 AI 图像识别、语音播报与模式切换的导盲助手应用，专为视障用户设计，助力他们更安全、自主地探索环境。

🌟 功能特点

• 📷 图像识别
  使用 YOLOv5 和 GPT-4o 模型识别物体并生成中文或英文描述

• 🗣️ 语音播报
  自动朗读识别内容，支持中英文语音提示

• 📍 定位支持
  根据模式决定是否启用 GPS 定位功能

• 🧭 三种模式支持
  三种使用模式优化语音交互与识别重点：

  • 户外模式：识别道路、障碍物、红绿灯等
  • 室内模式：识别洗手间、出入口、门等
  • 普通模式：常规识别与播报

🔧 技术栈

• 前端客户端：Basic4Android (B4A)
• 后端服务器：Flask + YOLOv5 + OpenAI GPT-4o
• 部署平台：支持本地与云部署（如 Render）

🚀 快速开始

📦 安装后端

git clone https://github.com/pandavio/vision-server.git
cd pdv-vision-server
pip install -r requirements.txt
python app.py

📱 部署客户端

• 使用 B4A 打开 PDVVision.b4a 项目
• 修改 Starter.ServerURL 为你的服务器地址
• 编译并安装至 Android 手机

🔑 所需 API 密钥

PDV Vision 的运行依赖以下两个 API 密钥：

• 🧠 OpenAI API 密钥：用于调用 GPT-4o 多模态模型，实现图像内容理解与自然语言描述
• 🗺️ Google Maps API 密钥：用于获取用户 GPS 定位与地址信息，提升户外导航体验（仅户外模式使用）

请在部署前完成以下设置：

• B4A 客户端：将 OpenAI 和 Google Maps 的 API 密钥配置在代码中相应位置
• Flask 后端：使用环境变量或配置文件导入 OpenAI API 密钥（如 OPENAI_API_KEY）

📂 项目结构

PDVVision/
├── B4A/              # Android 客户端代码
├── server/           # Flask 后端服务器
├── README.md
├── requirements.txt

📜 开源协议

本项目基于 MIT 协议开源。