---
title: SONAR-AI
emoji: 🔱
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.5.1
app_file: app.py
pinned: true
license: mit
short_description: X-Ray Cargo Analysis + 2128 HS Codes - 96% Accuracy
---

---

## 📊 Model Performance

| Metric | Value |
|--------|-------|
| **Accuracy** | **96.00%** |
| **Match Recall** | 100% |
| **Mismatch Precision** | 100% |
| **F1-Score** | 0.89 |
| **Parameters** | 5.67M |

---

## 🧠 Architecture: Deep-SOSUFS

**Deep-SOSUFS** (Self-Organizing Sonar Unsupervised Feature Selection):

- ✅ **EfficientNet-B0** backbone
- ✅ **Self-organizing clustering** layers
- ✅ **Learnable feature importance**
- ✅ **Attention mechanisms**

---

## 📋 Dataset

| Info | Value |
|------|-------|
| Total Images | 125 |
| Match | 112 (89.6%) |
| Mismatch | 13 (10.4%) |
| Source | Iraqi Customs |

---

## 🎯 Classification

- ✅ **Match**: Cargo matches declaration
- ❌ **Mismatch**: Cargo does NOT match declaration

---

## 👨‍🔬 Author

**Dr. Abbas Fadel Jassim Al-Jubouri**

- 🏛️ Iraqi General Customs Authority
- 🎓 Universiti Kebangsaan Malaysia (UKM)
- 📅 2025-2026

---

## 📚 Citation
```bibtex
@article{aljubouri2026deepsosufs,
  title={Deep-SOSUFS: Self-Organizing Feature Selection for X-Ray Cargo Analysis},
  author={Al-Jubouri, Abbas Fadel Jassim},
  year={2026},
  institution={Universiti Kebangsaan Malaysia}
}
```

---

**© 2026 Dr. Abbas Al-Jubouri | UKM Malaysia**