Simuletic/CCTV_Aggressive_Poses_Fight_Detection_Dataset

Fight Detection Dataset (CCTV, YOLO pose): free 103-image sample

This fight detection dataset is a free sample of 103 synthetic CCTV images of people fighting or about to fight, with YOLO pose labels (bounding box plus 17 body keypoints) for every person, for training and testing fight, violence and aggressive-pose detection models on security-camera footage. The full 1,000-image dataset adds a fight box around each group in contact, about 28% look-alike negatives (hugs, handshakes, dancing, arguments) and day, dusk, night and IR views, in YOLO-pose and COCO 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	103 JPEG	1,000 (900 train / 100 val)
Classes	person (one class)	person with 17 keypoints, plus fight boxes
Boxes	207 person, each with 17 keypoints	every person boxed; one fight box per group in contact
Hard negatives	none	about 28%: hugs, handshakes, high fives, dancing, arguments without contact
Resolution	784 × 1,168 to 1,206 × 1,560 px	CCTV frames; day, dusk, night and monochrome IR
Camera view	elevated CCTV angles, indoor and outdoor	streets, car parks, nightlife areas, stations, shops, lobbies, concourses
Labels	YOLO pose: box + 17 COCO keypoints	YOLO-pose + COCO, dataset card
Licence	CC BY 4.0	one-time: research $149 / commercial $590 per 1,000 images

What's in the images

People squaring up, shoving, punching or about to fight, seen from elevated CCTV cameras outside pubs, in car parks, shopping arcades, underpasses and streets, by day and at night. 102 images show two people and one shows three: 207 people in total. Every person has a bounding box and a 17-point body skeleton, so models can learn body language and pre-fight cues, not just where people are. In this sample every person has the same class, person.

Classes and boxes

Class	YOLO id	Boxes in sample
person	0	207

Label format and folder layout

Aggressive_Poses_Dataset/
  data.yaml                # class name and keypoint shape
  images/                  # 103 JPEG images (aggressive_pose_1.jpg, ...)
  labels/                  # one YOLO pose .txt per image

Each label line is one person: class_id x_center y_center width height, then 17 keypoints as x y v (56 values in total). Coordinates are normalised to 0–1. The keypoints follow the COCO order (nose, eyes, ears, shoulders, elbows, wrists, hips, knees, ankles); v is 2 for a labelled keypoint and 0 for one that is not labelled. Shortened example:

0.000000 0.310 0.675 0.379 0.644  0.391 0.427 2.000000  ...  0.180 0.938 2.000000

Values are written as decimals (0.000000 for class 0, 2.000000 for visible); Ultralytics reads them as normal.

Quick start (Ultralytics YOLO)

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

path: Aggressive_Poses_Dataset
train: images
val: images   # small sample: use the same images, or split them yourself
kpt_shape: [17, 3]
flip_idx: [0, 2, 1, 4, 3, 6, 5, 8, 7, 10, 9, 12, 11, 14, 13, 16, 15]
names:
  0: person

Then train:

from ultralytics import YOLO

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

Use cases

• Fight and violence detection on CCTV and video management systems (VMS)
• Early warning of confrontations in nightlife venues, public transport and schools
• Pose-based action recognition and body-language research
• Pre-training before fine-tuning on your own camera footage

Full dataset and prices

The full CCTV Fight & Aggression Detection Dataset has 1,000 images (900 train / 100 val), YOLO-pose + COCO, person boxes with 17 keypoints plus fight boxes, about 28% look-alike negatives. 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 labels everyone as person; the full dataset separates normal_pose from aggressive_pose.

More synthetic datasets from Simuletic

• CCTV Fall & Incident Detection Dataset
• CCTV Knife Detection Dataset
• CCTV Weapon Detection Dataset
• CCTV Pedestrian-1K: Person Attribute Dataset
• 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_fight_2025,
  author       = {{Simuletic}},
  title        = {CCTV Fight & Aggression Detection Dataset},
  year         = {2025},
  publisher    = {Simuletic},
  howpublished = {\url{https://huggingface.co/datasets/Simuletic/CCTV_Aggressive_Poses_Fight_Detection_Dataset}},
  note         = {Free sample. Full dataset: \url{https://simuletic.com/datasets/cctv-aggressive-pose-detection-dataset}}
}

Questions or feedback: contact@simuletic.com