---
title: Multilingual Image Describer
emoji: 🌍
colorFrom: blue
colorTo: purple
sdk: streamlit
sdk_version: "1.28.1"
app_file: app.py
pinned: false
---

# 🌍 Multilingual Image Describer with Real Translation

A powerful web application that analyzes images to detect objects and generate descriptions in **14 different languages** with real translation capabilities.

## ✨ Features
- **Real-time Translation**: Uses Hugging Face NLLB model for accurate translations
- **Object Detection**: Powered by YOLOv8 with confidence scores
- **Image Captioning**: Generates natural language descriptions using BLIP
- **14 Languages**: English, Spanish, French, German, Chinese, Hindi, Arabic, Russian, Japanese, Korean, Portuguese, Italian, Amharic, Turkish
- **Export Results**: Download analysis as JSON or TXT files

## 🚀 How to Use

### 1. Upload Image
- Click "Upload" to select an image file
- Use "Camera" to capture a photo
- Try "Sample" for a demo image

### 2. Configure Settings
- **Select Language**: Choose from 14 available languages
- **API Token**: Optional Hugging Face token for better translation
- **Confidence**: Adjust object detection sensitivity

### 3. Analyze
- Click "🚀 Analyze Image" to process
- View results in Description, Objects, and Export tabs
- Download results for later use

## 🔧 Translation Setup

For best translation results:

1. **Get a free Hugging Face token**:
   - Visit [huggingface.co/settings/tokens](https://huggingface.co/settings/tokens)
   - Create a new token (select "Read" permission)
   - Copy the token (starts with `hf_`)

2. **Enter token in the app**:
   - Paste token in the "Hugging Face Token" field in sidebar
   - Token is stored locally for your session

## 🌐 Supported Languages

| Language | Code | Support |
|----------|------|---------|
| English | en | ✅ Full |
| Spanish | es | ✅ Full |
| French | fr | ✅ Full |
| German | de | ✅ Full |
| Chinese | zh | ✅ Full |
| Hindi | hi | ✅ Full |
| Arabic | ar | ✅ Full |
| Russian | ru | ✅ Full |
| Japanese | ja | ✅ Full |
| Korean | ko | ✅ Full |
| Portuguese | pt | ✅ Full |
| Italian | it | ✅ Full |
| Amharic | am | ✅ Full |
| Turkish | tr | ✅ Full |

## 📊 Technical Details

- **Image Models**: BLIP (captioning) + YOLOv8 (detection)
- **Translation**: Facebook NLLB-200-distilled-600M via Hugging Face Inference API
- **Framework**: Streamlit + Python 3.10
- **Deployment**: Hugging Face Spaces

## 🛠️ Development

### Local Setup
```bash
git clone https://huggingface.co/spaces/amogneandualem/amogne-vlm-LLM
cd amogne-vlm-LLM
pip install -r requirements.txt
streamlit run app.py