Simuletic/CCTV_Weapon_Detection_Rifles_vs_Umbrellas

Synthetic Weapon Detection: Rifles vs. Umbrellas (CCTV)

Overview

This is an open-source synthetic dataset designed to solve the single biggest problem in Weapon Detection AI: False Positives.

Standard models often confuse common handheld objects—like closed umbrellas, tripods, or tools—with firearms. This dataset focuses specifically on Hard Negative Mining, containing a balanced split of lethal weapons (Rifles) and visual lookalikes (Umbrellas) seen from realistic CCTV angles.

🚀 Need more data? This is a sample dataset by Simuletic. We generate hyper-realistic synthetic data to fix edge cases and bias in AI training. We have a larger dataset with 1000+ images Handguns, Drills, and Crowds, visit simuletic.com/datasets.

Key Features

• Hard Negative Focus: ~50% Rifles (Threat) vs. ~50% Umbrellas (Confuser). Train your model to distinguish the subtle difference in grip and posture.
• CCTV Realism: High-angle, overhead perspectives with realistic sensor noise, motion blur, and varied lighting conditions.
• Privacy-First: Fully synthetic. No real individuals are depicted, eliminating GDPR and privacy risks.
• YOLO Ready: Pre-annotated in standard YOLO TXT format, ready for immediate training with YOLOv8, v10, or v11.

Dataset Structure & Classes

The dataset follows the standard YOLO format structure.

• images/: High-fidelity synthetic .jpg files.
• labels/: .txt files containing class ID and bounding box coordinates.

Class Map

• 0: person (A person)
• 1: rifle (Assault rifles, carbines held in various postures)
• 2: umbrella (Closed black umbrellas, held by stem or handle)

YAML Configuration

To use this with Ultralytics YOLO, your data.yaml should look like this:

path: /path/to/dataset
train: images
val: images

nc: 3
names:
  0: person
  1: rifle
  2: umbrella
Use Cases
Active Shooter Detection: Improve the precision of security systems in schools, airports, and public spaces.

False Alarm Reduction: Drastically reduce the number of false alerts caused by innocent daily objects.

Perimeter Security: Train models to ignore non-threat objects carried by staff or visitors.

## Ethics & License
Synthetic Nature: This data is 100% computer-generated. No real humans were recorded, staged, or harmed in the making of this dataset.

## License: CC BY 4.0. 
You are free to use and adapt this data for research or commercial projects, provided you give appropriate credit to Simuletic.

Citation
If you use this dataset in your research or projects, please cite:

Kodavsnitt
@dataset{simuletic_weapon_umbrella_2026,
  author = {Simuletic Team},
  title = {Simuletic Synthetic Weapon vs. Umbrella Detection Dataset},
  year = {2026},
  url = {[https://simuletic.com](https://simuletic.com)}
}

## Feedback? Reach out via simuletic.com or the "Issues" tab here on Kaggle.