UXO-Politehnica-Bucharest/Contextual_Vision_for_Unexploded_Ordnances

🧨 CTX-UXO: Contextual Vision Dataset for Unexploded Ordnances

📖 Description

According to US NOAA, unexploded ordnances (UXO) are “explosive weapons such as bombs, bullets, shells, grenades, mines, etc. that did not explode when they were employed and still pose a risk of detonation”. UXOs are among the most dangerous threats to human life, environment and wildlife protection as well as economic development. The risks associated with UXOs do not discriminate based on age, gender, or occupation, posing a danger to anyone unfortunate enough to encounter them.

Contrary to expectations, an UXO is more hazardous than new ordnance, as its arming or initiation mechanisms may be active or compromised. A mistake in correctly identifying ordnance can be fatal, which is why a decision support system can assist in making decisions under continuous stress, where lives are at risk.

Recent advances in image processing techniques and deep learning demonstrate that object detection and identification can be applied across multiple domains. However, until now, UXO detection has been limited by the lack of a representative, comprehensive dataset that provides robustness across different scenarios. UXOs are often found in altered, oxidized, semi-buried states in hard-to-reach environments.

---

🧠 CTX-UXO Dataset

We thus propose the Contextual Vision for Unexploded Ordnances (CTX-UXO) dataset, which provides a collection of labeled UXO images in various visual contexts within the visible spectrum.

The dataset encompasses ammunitions in different stages, across multiple environments, angles, distances, and with various types of cameras. Additionally, replicas of munitions, faithfully replicating the characteristics of real ordnance, were used to diversify the dataset by relocating or arranging them in new positions and environments, and by removing certain ordnance components.

This approach aims to create a dataset that is as varied and representative of real-world scenarios as possible. The dataset will be periodically updated with new types of UXO in different visual contexts.

We hope that this dataset represents a useful resource for researchers and engineers working on supervised and semi-supervised object recognition projects, with particular emphasis on civil protection and emergency situation management applications.

---

✅ Validation and Publications

Undergoing: Data Descriptor Paper

The CTX-UXO dataset has been rigorously validated using a wide range of deep learning architectures and methodological frameworks, incorporating diverse preprocessing techniques. The resulting findings have been disseminated through peer-reviewed scientific publications:

• [1] Craioveanu M., Stamatescu I., Stamatescu G.
  Ensemble Strategy With Multi-Step Hard Sample Mining for Improved UXO Localisation and Classification,
  IEEE Access, vol. 13, pp. 123546–123558, 2025.

• [2] Craioveanu M., Stamatescu G.
  Detection and Identification of Unexploded Ordnance using a Two-Step Deep Learning Methodology,
  32nd Mediterranean Conference on Control and Automation (MED 2024), June 11–14, Chania, Greece.

• [3] Craioveanu M., Stamatescu G.
  Evaluation of the Robustness–Runtime Efficiency Trade-Off of Edge AI Models in UXO Localisation and Classification,
  33rd Mediterranean Conference on Control and Automation (MED 2025), June 10–13, Tangier, Morocco.

---

🙏 Acknowledgements

We would like to thank the personnel of the National Romanian Inspectorate for Emergency Situations for their logistical support in the collection and dissemination of this dataset.

---

📁 Dataset Structure

• Images: 3,520 JPG files
• Instances: 15,449
• Average instances/image: 4.38
• Image format: RGB (.jpg)
• Median resolution: ~2124×2124 px
• Splits:
  • train (70%)
  • validation (15%)
  • test (15%)
    (using Multilabel Stratified Shuffle Split)

---

📷 Image Acquisition

Captured using various high-performance devices, predominantly mobile phones.

• Most frequent camera: 64MP model with Samsung GW3 sensor (S5KGW3)
  • 1/1.97" imager
  • 0.7µm pixels
  • Tetra-cell technology
  • 25mm f/1.8 lens

---

🧾 Annotations

• Supported formats: YOLO and COCO
• Organized into multiple repositories for:
  • Binary classification
  • Multi-class detection
  • Instance segmentation

Future versions will include more UXO types under diverse lighting and environmental conditions.

---

📊 Class Distribution

Class	Instances
Projectile	6,121
Mortar Bomb	4,269
Grenade	3,399
Cartridge	987
Aviation Bomb	333
Cartridge Magazine	124
Fuse	92
RPG	63
Landmine	29
Rocket	21
AntiSubmarine	6
Sea Mine	5

---

🧪 Binary vs. Multi-Class Use

• Binary classification (UXO / Non-UXO): full 15,449 instances
• Multi-class detection and classification: use per-class labels (e.g., Mortar Bomb, Projectile, Grenade, etc.)

---

📬 Contact

For questions or collaborations, feel free to contact the authors via institutional emails or through the Hugging Face community discussion tab.

---