Behaviour & safetyCCTV Shoplifting Detection
Train a shoplifting detection model: 5,000+ synthetic CCTV frames and 100+ videos with YOLO, pose and VLM labels. Free 400-image sample on Kaggle.
- Size
- 5,000+ frames + 100+ videos
- Format
- YOLO + VLM
CCTV datasets
CCTV datasets for computer vision
Security-camera training data seen from where the cameras hang: ceiling, wall and pole mounts, day, night and infrared, with the look-alikes that cause false alarms.
9 datasets
Behaviour & safety
Weapons & threatsCCTV ATM Robbery Detection: Gun & Knife
Synthetic ATM robbery detection dataset with 3,000 YOLO images. Four classes: offender, victim, gun, knife. 237 open-source samples on Kaggle.
- Size
- 3,000 images
- Format
- YOLO
Fire & smokeCCTV Smoke & Fire Emergency Detection
Smoke and fire detection dataset: 1,000 synthetic CCTV images of early-stage fires plus look-alike negatives, person/smoke/fire boxes in YOLO and COCO.
- Size
- 1,000 images
- Format
- YOLO + COCO
Weapons & threatsCCTV Weapon Detection
Train AI weapon detection on 1,000 synthetic CCTV images: handguns, rifles, shotguns and look-alike negatives, labelled in YOLO and COCO. Free sample included.
- Size
- 1,000 images
- Format
- YOLO + COCO
Weapons & threatsCCTV Knife Detection
Synthetic knife detection dataset: 1,000 CCTV images of people holding knives plus look-alike negatives, labelled in YOLO and COCO. Free sample.
- Size
- 1,000 images
- Format
- YOLO + COCO
Behaviour & safetyCCTV Fall & Incident Detection
Synthetic fall detection dataset: 1,000 CCTV images of fallen people plus sitting and kneeling look-alikes, labelled person / fallen_person in YOLO and COCO.
- Size
- 1,000 images
- Format
- YOLO + COCO
Behaviour & safetyCCTV Fight & Aggression Detection
Fight detection dataset: 1,000 synthetic CCTV images with person boxes, 17-keypoint poses and fight boxes in YOLO-pose and COCO, plus look-alike negatives.
- Size
- 1,000 images
- Format
- YOLO-pose + COCO
Behaviour & safetyPedestrian Attribute Recognition
Synthetic pedestrian attribute dataset — a PA-100k alternative with captions and structured labels for ViT, CLIP and Re-ID. Free 996-image sample.
- Size
- 10,000 images (50K by quote)
- Format
- JSONL (Descriptive + Structured)
Weapons & threatsCCTV Weapon Detection: Rifles vs Umbrellas
Synthetic hard-negative dataset: rifles vs umbrellas to cut weapon-detection false positives. 1,000 YOLO images. Open-source sample available.
- Size
- 1,000 images
- Format
- YOLO
Public CCTV datasets are either small academic sets filmed by one group (UCF-Crime, CAUCAFall, PoseLift) or scraped footage with unclear rights. Both are hard to license for a product. Every dataset on this page is synthetic: rendered from CCTV camera positions, so there are no real people, no consent to clear and no footage you cannot redistribute as a trained model.
Each set ships with YOLO and COCO boxes, a fixed train and validation split, a dataset card, and a free sample on Hugging Face or Kaggle so you can check the labels before paying. Hard negatives (umbrellas for rifles, phones for knives, people sitting for falls, steam for smoke) are kept in every set in a known share.
Licences are one-time and per dataset: $149 research or $590 commercial for 1,000 images, and packs of 5,000–20,000 images at $2,400–$8,500 commercial. The shoplifting set is video and is priced on request.
Prices
One-time licences, per dataset.
| Images | Research | Commercial | Order |
|---|---|---|---|
| 1,000 | $149 | $590 | On the dataset page |
| 5,000 | Not offered | $2,400 | Order 5,000 |
| 10,000 | Not offered | $4,500 | Order 10,000 |
| 20,000 | Not offered | $8,500 | Order 20,000 |
- Commercial licence only: the research licence covers 1,000 images
- YOLO + COCO labels, fixed train / val split, dataset card
- Hard negatives in the same share as the 1,000-image set
- Masks, pose keypoints or VLM captions where the dataset has them
- Card or invoice; delivered as a download link within five working days
Larger packs are assembled for your order and delivered as a download link within five working days. Prices in USD, excl. VAT, one-time, per dataset.
Read next
Guides for this kind of data.
Guide
CCTV and Surveillance Datasets: Public Sources, Licences and Gaps (2026)
17 public CCTV and surveillance datasets compared by task, size, camera, labels and licence, from UCF-Crime to MOT20, and which ones allow commercial use.
Guide
AI Security Camera Training Data: How Detectors Are Trained
Buyer's checklist for AI security camera training data: classes and hard negatives, YOLO/COCO labels, camera geometry, licences, validation and the EU AI Act.
Guide
CCTV Video Anomaly Detection Datasets: A Practical Guide
How CCTV video anomaly detection datasets are built, which behaviors matter (falls, fights, weapon draws, loitering), and why synthetic CCTV wins.
Questions
CCTV datasets: common questions
What is a CCTV dataset?
A collection of images or video clips taken, or rendered, from security-camera positions and labelled for a computer-vision task such as weapon detection, fall detection or person attributes. Labels are usually bounding boxes in YOLO or COCO format.
Where can I download a CCTV dataset for free?
Every Simuletic CCTV dataset has a free sample of 100 to 400 labelled images on Hugging Face or Kaggle. Public academic sets such as UCF-Crime, CAUCAFall and RWF-2000 are free for research but mostly exclude commercial use.
Can I use a CCTV dataset commercially?
Synthetic sets can be: Simuletic's commercial licence ($590 per 1,000 images, one-time) lets one organisation train and ship models. Real-footage academic datasets are usually research-only, and scraped CCTV raises GDPR questions regardless of the licence.
Are the datasets real CCTV footage?
No. Every frame is generated from a CCTV camera specification (height, tilt, lens, lighting, compression), so the geometry matches real cameras while no real person appears.
More collections
Other ways into the catalogue.
Show us the alarm your model keeps getting wrong.
Send five frames from your cameras. We reply within one working day with the fastest way to fix it — a catalogue dataset, a stress test or a custom pack.