Simuletic/cctv-weapon-dataset

Weapon Detection Dataset (CCTV, YOLO): free 141-image sample

This weapon detection dataset is a free sample of 141 synthetic CCTV images of people holding guns, with YOLO bounding boxes for person and weapon, for training and testing weapon and gun detection models on security-camera footage. The full 1,000-image dataset separates handgun, rifle, shotgun and sniper_rifle, adds COCO labels and about 20% look-alike hard negatives, and ships with a 900 / 100 train-val split. Made by Simuletic, a Swedish company that has built synthetic CCTV datasets for computer vision since 2023.

Get the full dataset · Free sample licence: CC BY 4.0. This free sample is also on Kaggle and GitHub.

Facts

	Free sample (this repo)	Full dataset
Images	141 PNG	1,000 JPEG (900 train / 100 val)
Classes	person, weapon	person, handgun, rifle, shotgun, sniper_rifle
Boxes	139 person, 130 weapon	every person and weapon boxed
Hard negatives	none	about 20%: bottles, umbrellas, bags, phones, tools
Resolution	about 1,100 × 1,100 to 924 × 1,358 px	CCTV frames; day, night and monochrome IR
Camera view	elevated CCTV angles, indoor and outdoor	ceiling, wall and pole mounts; shops, offices, stations, warehouses, streets
Labels	YOLO .txt + one CSV of pixel boxes	YOLO + COCO, dataset card
Licence	CC BY 4.0	one-time: research $149 / commercial $590 per 1,000 images

What's in the images

People holding or carrying firearms in public and semi-public places (entrances, corridors, streets, car parks), seen from elevated security-camera positions. Weapons are often small in the frame, as they are in real CCTV footage. The sample comes from an earlier Simuletic release and uses two classes; the full dataset separates weapon types and adds hard negatives.

Classes and boxes

Class	YOLO id	Boxes in sample
person	0	139
weapon	1	130

Label format and folder layout

samples/
  images/                  # 141 PNG images (Scene1_1.png, ...)
  labels/                  # one YOLO .txt per image
  annotation_sample.csv    # the same boxes in pixels

Each YOLO label line is class_id x_center y_center width height, normalised to 0–1:

0 0.452 0.551 0.203 0.302
1 0.601 0.703 0.102 0.151

The CSV has one row per box: label_name, bbox_x, bbox_y, bbox_width, bbox_height, image_name, image_width, image_height (pixels, top-left origin).

Quick start (Ultralytics YOLO)

Create a data.yaml next to the samples/ folder:

path: samples
train: images
val: images   # small sample: use the same images, or split them yourself
names:
  0: person
  1: weapon

Then train:

from ultralytics import YOLO

model = YOLO("yolo11n.pt")
model.train(data="data.yaml", epochs=30, imgsz=640)

A Colab notebook that fine-tunes YOLO11 on the newer 5-class sample is in the GitHub repo.

Use cases

• Gun and weapon detection on CCTV and video management systems (VMS)
• Real-time threat detection in shops, offices, stations and public spaces
• Pre-training before fine-tuning on your own camera footage
• Benchmarking detectors on small, distant objects

Full dataset and prices

The full CCTV Weapon Detection Dataset has 1,000 images, YOLO + COCO, classes person, handgun, rifle, shotgun, sniper_rifle. Every licence is a one-time payment, perpetual, for one organisation; you get a download link within five working days (card or invoice). Research covers students and non-commercial work and is available for 1,000 images only; packs of 5,000 images and more are sold with the commercial licence, which lets you train models and ship them in products.

	Research	Commercial
1,000 images	$149	$590
5,000-image pack	not offered	$2,400
10,000-image pack	not offered	$4,500
20,000-image pack	not offered	$8,500

Packs are the same dataset at larger size, with the same labels and split, under the commercial licence. 50,000+ images, video datasets and enterprise agreements (from $10,000) are quoted. Other scenes, camera angles or classes: custom data or contact@simuletic.com.

How the images were made

The images are fully synthetic (computer-generated) by Simuletic: no real people, places or incidents are depicted, so there are no privacy or consent issues.

Limitations. Synthetic images do not cover every camera, lens, compression artefact or scene you will meet in deployment. Validate on your own footage, and for production combine synthetic and real data. This sample has one generic weapon class and no hard negatives; the full dataset has both.

More synthetic datasets from Simuletic

• CCTV Knife Detection Dataset
• CCTV Weapon Detection: Rifles vs Umbrellas
• Weapon Detection: Handgun vs Bag of Chips
• CCTV ATM Robbery Detection Dataset: Gun & Knife
• All datasets: simuletic.com/datasets · Hugging Face · Kaggle · GitHub

Licence

This free sample is licensed under CC BY 4.0: you may share and adapt it for any purpose, including commercial use, as long as you credit Simuletic. The full dataset is sold under a separate one-time research or commercial licence; it is not redistributed.

Citation

@misc{simuletic_weapon_2025,
  author       = {{Simuletic}},
  title        = {CCTV Weapon Detection Dataset},
  year         = {2025},
  publisher    = {Simuletic},
  howpublished = {\url{https://huggingface.co/datasets/Simuletic/cctv-weapon-dataset}},
  note         = {Free sample. Full dataset: \url{https://simuletic.com/datasets/cctv-weapon-detection-dataset}}
}

Questions or feedback: contact@simuletic.com