🪔 DARA — Detect & Assist Recognition AI

"Mata untuk semua" (Eyes for everyone)

🇺🇸 English | 🇮🇩 Bahasa Indonesia

---

English

🎯 Mission

DARA is an open-source, lightweight Vision Language Model (VLM) designed for assistive technology. It provides a "second pair of eyes" for visually impaired individuals, elderly users, and the general public.

Key Features:

• 🚀 Ultra-fast: 300-500ms inference on CPU
• 📱 Mobile-ready: Runs on phones and edge devices
• 🧠 5 Smart Modes: Scene, Emotion, Medicine, Currency, Text
• 🔊 Voice Output: Integrated TTS for accessibility
• 🌐 Offline: Works without internet connection
• 🇮🇩 Bilingual: English and Indonesian support

🌟 The 5 Smart Modes

Mode	Icon	Function	Example Output
Scene	🏞️	Describes environment + hazard detection	"Kitchen with table. ⚠️ Stove is on."
Emotion	😊	Reads facial expressions + social guidance	"Happy. They seem approachable."
Medicine	💊	Reads labels + dosage extraction	"Paracetamol 500mg. Take after meals."
Currency	💵	IDR detection with colors	"Rp 50.000 (blue color)"
Text	📝	OCR for any text	"Exit sign. Door to your left."

🚀 Quick Start

# Clone repository
git clone https://github.com/ardelyo/dara.git
cd dara/dara_project

# Install dependencies
pip install -r requirements.txt
pip install -e .

# Run demo
python app.py

💻 Basic Usage

from dara import DARA

# Initialize
dara = DARA()

# Detect with mode  
result = dara.detect("photo.jpg", mode="scene", language="en")

print(result["result"])      # "Modern kitchen with table..."
print(result["confidence"])  # 0.85
print(result["audio"])       # Path to TTS audio

📊 Performance

Metric	Value
Import Time	~16s (first load)
Inference (CPU)	300-500ms
Cache Hit	<1ms
Memory	~1.5GB

---

Bahasa Indonesia

🎯 Misi

DARA adalah Model Vision-Language (VLM) open-source yang ringan, dirancang untuk teknologi asistif. Menyediakan "sepasang mata kedua" untuk tunanetra, lansia, dan masyarakat umum.

Fitur Utama:

• 🚀 Ultra-cepat: Inferensi 300-500ms di CPU
• 📱 Siap Mobile: Berjalan di HP dan edge device
• 🧠 5 Mode Cerdas: Scene, Emosi, Obat, Mata Uang, Teks
• 🔊 Output Suara: TTS terintegrasi untuk aksesibilitas
• 🌐 Offline: Bekerja tanpa koneksi internet
• 🇮🇩 Bilingual: Dukungan Inggris dan Indonesia

🌟 5 Mode Cerdas

Mode	Ikon	Fungsi	Contoh Output
Scene	🏞️	Deskripsi lingkungan + deteksi bahaya	"Dapur dengan meja. ⚠️ Kompor menyala."
Emotion	😊	Baca ekspresi + saran sosial	"Senang. Terlihat ramah."
Medicine	💊	Baca label + ekstraksi dosis	"Paracetamol 500mg. Minum setelah makan."
Currency	💵	Deteksi Rupiah dengan warna	"Rp 50.000 (warna biru)"
Text	📝	OCR untuk teks apapun	"Tanda keluar. Pintu di kiri."

🚀 Mulai Cepat

# Clone repository
git clone https://github.com/ardelyo/dara.git
cd dara/dara_project

# Install dependencies
pip install -r requirements.txt
pip install -e .

# Jalankan demo
python app.py

💻 Penggunaan Dasar

from dara import DARA

# Inisialisasi
dara = DARA()

# Deteksi dengan mode
result = dara.detect("foto.jpg", mode="scene", language="id")

print(result["result"])      # "Dapur modern dengan meja..."
print(result["confidence"])  # 0.85
print(result["audio"])       # Path ke audio TTS

📊 Performa

Metrik	Nilai
Waktu Import	~16 detik (muat pertama)
Inferensi (CPU)	300-500ms
Cache Hit	<1ms
Memori	~1.5GB

---

📁 Project Structure | Struktur Proyek

dara_project/
├── src/dara/              # Main package
│   ├── core/              # Model, processor, inference
│   ├── modes/             # 5 mode handlers
│   ├── services/          # TTS, translation, cache
│   ├── utils/             # Utilities
│   ├── config.py          # Configuration
│   └── __init__.py        # Public API
├── demo/                  # Demo applications
├── scripts/               # Training & benchmark
├── docs/                  # Documentation
│   ├── QUICKSTART.md      # ⭐ Start here!
│   ├── API.md             # API reference
│   ├── ARCHITECTURE.md    # System design
│   ├── TRAINING.md        # Training guide
│   ├── RESEARCH.md        # Research paper
│   ├── STATISTICS.md      # Benchmark data
│   └── CONTEXT.md         # Vision & roadmap
├── data/                  # Training data
└── tests/                 # Unit tests

📚 Documentation | Dokumentasi

Document	Description	Deskripsi
QUICKSTART.md	Quick start guide	Panduan mulai cepat
API.md	API reference	Referensi API
ARCHITECTURE.md	System design	Desain sistem
TRAINING.md	Training guide	Panduan training
RESEARCH.md	Research paper	Makalah penelitian
STATISTICS.md	Benchmark data	Data benchmark
CONTEXT.md	Vision & roadmap	Visi & roadmap

🔢 Version History | Riwayat Versi

Version	Date	Changes
0.2.0	Dec 2024	Modular architecture, caching, bilingual docs
0.1.0	Nov 2024	Initial release with 5 modes

🤝 Contributing | Kontribusi

Contributions welcome! See CONTRIBUTING.md for guidelines.

Kontribusi dipersilakan! Lihat CONTRIBUTING.md untuk panduan.

📄 License | Lisensi

Apache 2.0 - See LICENSE for details.

---

Built with ❤️ for accessibility | Dibangun dengan ❤️ untuk aksesibilitas

"Mata untuk semua" | "Eyes for everyone"