visuAI-tensorflow

Intelligent Image Recognition with OmniRL Vision-Language Understanding

📋 Overview

visuAI is an advanced image recognition system that combines:

• Frontend: Angular app with TensorFlow.js (MobileNet) for fast classification
• Backend: FastAPI + OmniRL for intelligent descriptions and Q&A

🚀 Quick Start

Frontend Setup

1. Install dependencies:

npm install

2. Run development server:

ng serve

Visit http://localhost:4200

Backend Setup

1. Navigate to backend:

cd backend

2. Create virtual environment:

python -m venv venv
venv\Scripts\activate  # Windows

3. Install dependencies:

pip install -r requirements.txt

4. Configure environment:

copy .env.example .env

5. Run server:

python main.py

API available at http://localhost:8000

🏗️ Architecture

Image Upload → MobileNet (Browser) → Predictions
                                          ↓
                                    FastAPI Backend
                                          ↓
                                    OmniRL Model
                                          ↓
                              Description + Q&A ← User

✨ Features

Frontend

• Fast Classification: TensorFlow.js runs in browser (no server needed)
• Real-time Results: Instant prediction probabilities
• Modern UI: Angular Material design

Backend

• Smart Descriptions: Converts predictions to natural language
• Visual Q&A: Answer questions about images
• Caching: Fast responses for repeated queries

📖 Documentation

• Frontend: See Angular docs
• Backend API: http://localhost:8000/docs (when running)
• Implementation Plan: See project artifacts

🔧 Tech Stack

• Frontend: Angular 18, TensorFlow.js, Material UI
• Backend: Python, FastAPI, PyTorch (OmniRL)
• ML Models: MobileNet (classification), Qwen2.5-VL-3B (VQA)

📂 Project Structure

visuAI-tensorflow/
├── src/                    # Angular frontend
│   ├── app/
│   │   ├── components/
│   │   └── services/
│   └── assets/
├── backend/               # FastAPI backend
│   ├── main.py
│   ├── models/
│   ├── services/
│   └── training/
└── ...

🎯 Current Status

✅ Phase 1: Backend structure complete (mock mode)
⏳ Phase 2: OmniRL training in progress
⏳ Phase 3: Frontend integration

📝 License

MIT License

🤝 Contributing

Contributions welcome! This is an experimental project for vision-language integration.