README.md

metadata

title: Advanced Multilingual Image Describer
emoji: 🌍
colorFrom: purple
colorTo: indigo
sdk: streamlit
sdk_version: 1.32.0
app_file: app.py
pinned: false

🌍 Advanced Multilingual Image Describer

No translation APIs • Native multilingual support • Latest vision-language models

🚀 Features

• Direct multilingual captioning - No separate translation step
• Latest models - LLaVA 1.5, Qwen-VL, Moondream 2
• 10+ languages - Native support for English, Chinese, Amharic, Spanish, French, German, Arabic, and more
• Fast & efficient - Optimized for Hugging Face Spaces
• Clean interface - Simple and intuitive

🤖 Supported Models

LLaVA 1.5 (7B)

• Languages: English, Chinese, Spanish, French, German, Italian, Russian, Japanese, Korean, Arabic
• Best for: High-quality detailed descriptions
• Size: 7 billion parameters

Qwen-VL-Chat

• Languages: English, Chinese, Japanese, Korean, French, German, Spanish, Russian
• Best for: Conversational responses
• Size: 9.6 billion parameters

Moondream 2

• Languages: English, Spanish, French, German
• Best for: Fast inference, smaller size
• Size: 1.4 billion parameters

🌐 How It Works

1. Select a model from the sidebar
2. Choose language for output
3. Upload an image (JPG, PNG, WebP)
4. Click "Generate Description"
5. Get native description in selected language

⚡ Performance

• Inference time: 2-10 seconds
• Memory usage: ~8-16GB RAM
• Quality: Human-like descriptions
• Languages: Native output (not translated)

🛠️ Technical Details

• Framework: Streamlit + Transformers
• Models: Latest vision-language models from Hugging Face
• Deployment: Hugging Face Spaces (CPU/GPU)
• Code: Pure Python, no external APIs

📋 File Structure