01
Browse and buy a dataset
16 ready-made datasets with sample images. One-time research or commercial licences.
Security AI Datasets Built for Real-World Edge Cases
Licensed, privacy-safe CCTV datasets and training data for AI security cameras: weapon detection, incident detection, Counter-UAS, driver monitoring and other hard visual AI problems. QA-verified YOLO and COCO labels, hard negatives that cut false alarms, real CCTV angles and lighting — as images and video, or as a custom dataset built on your own cameras.
One empty scene in, labelled edge cases out
Customer cameraNo detection
Generated · knifeKnife
Generated · handgunWeapon
Each clip, with what it shows and which dataset it comes from, is on the sample videos pages.
- Ready-made datasets
- 16
- Label formats
- YOLO · COCO
- Real people in frame
- 0
- Licence per dataset
- One-time
Programmes
Claude for Startups · Databricks partner · Snowflake partner
01Get started
Three ways to start today.
Buy a ready-made dataset, have us build data on your own cameras, or join the API early-access list.
02
API coming soon: get early access
Generate and annotate labelled images from code or from an AI agent. Tell us what you'd build and we'll invite you to the beta.
03
Request a custom dataset
Images or video clips on your cameras, with your classes. Images from $1,500 for 2,500 images; video quoted per project.
02What you can buy
Three ways in. Pick the one that matches your deadline.
Most teams start with a dataset or a stress test, then move to a custom pack for their own cameras. Datasets and services are one-time purchases.
Ready today
Dataset library
One-timelicence per dataset
Licence one of 16 catalogue datasets, one-time. Free sample first, so you can check the labels before paying.
- 1,000 images · research or commercial licence
- YOLO + COCO, fixed train / val split
- 5,000 / 10,000 images on request
5 working days
False-Positive Stress Test
$1,500
We run your detector on our hard negatives and edge cases. You get every failure, ranked, with the frames.
- Umbrellas vs rifles, phones vs handguns, night, rain
- Precision / recall per scenario
- Fix list you can act on
Your cameras
Custom pack
from $1,500
Send 5–20 photos of your empty scenes. We return labelled images and video clips shot from exactly those angles.
- 2,500 images on your cameras
- Video clips with tracking IDs — quoted
- Your classes plus hard negatives
Building agents or pipelines? Get early access to the API and MCP server.
03Why it works
False alarms come from things that look like threats. So we render those too.
An umbrella carried like a rifle. A bag of chips held like a handgun. Every weapon dataset ships with the look-alikes that trigger real-world false positives, labelled as what they are.
Labels you can audit
Every dataset card states how the boxes were made and the measured label accuracy. Download the free sample and check it yourself.
Angles that match yours
High, wide CCTV views with compression, noise and low light — not stock photos shot at eye level.
No people, no GDPR exposure
Every frame is generated. There are no real faces or bystanders, so the data clears legal review fast.
04How custom data is made
From one photo of an empty scene to a labelled training set.
The same pipeline builds our catalogue and the custom datasets we make on your cameras.
Generated from one frameThreat
- 01
Send one clean frame of your scene
A hospital corridor, a station platform, a store entrance — empty, so no real people are involved from the start.
- 02
Say what should happen in it
People walking, running, falling, drawing a weapon, leaving a bag. Plus the look-alikes your model confuses today.
- 03
We render the variations
Different people, clothing, poses, lighting and weather in your exact view — as images, video clips, or both.
- 04
Every frame is labelled and checked
Boxes, classes and tracking IDs, exported as YOLO or COCO, with a QA pass before anything is delivered.
05Dataset library
Start with a free sample.
Weapons, behaviour, fire and counter-UAS scenarios. Each has a free sample on Kaggle and a one-time licence.
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 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
Behaviour & safetyCCTV Shoplifting Detection
Synthetic shoplifting detection dataset: 5,000+ CCTV frames, 100+ videos, YOLO + pose + VLM captions. 400 images and 8 videos open-source on Kaggle.
- Size
- 5,000+ frames + 100+ videos
- Format
- YOLO + VLM
Counter-UAS & defenceMilitary Drone Swarm & Saturation Attack
Synthetic drone swarm dataset for Counter-UAS training. 5-15 drones per frame, 10-50 pixel objects, adverse weather. 1,000+ YOLO-annotated images.
- Size
- 1,000+ images
- Format
- YOLO
06Questions
Frequently asked questions
Synthetic data, privacy, samples and pricing.
What is synthetic data for computer vision?
Synthetic data is imagery and video generated by rendering and generative models instead of being filmed. It is used to train and test detection models on events that are rare, dangerous or private — weapon draws, falls, fights, fire — without collecting real footage. Simuletic builds synthetic CCTV, drone and dashcam data for security, defence and safety teams.
How does synthetic data compare to real-world data?
Labels come from the pipeline and are checked before delivery, edge cases and look-alikes can be produced on demand, and there are no real people in frame. Most teams combine it with the real footage they already have: synthetic data fills the gaps — rare events, night, rain, unusual angles — that real collections rarely cover.
Which industries use Simuletic data?
Video-analytics vendors working on weapon and intrusion detection, transport and retail security, healthcare fall detection, industrial safety, smart-city monitoring, driver monitoring, and defence and counter-UAS teams. See the use-case overview for the datasets that fit each one.
Is synthetic data GDPR compliant?
Every frame is computer-generated, so there are no real individuals in the data. That removes consent requirements and most of the data-protection risk that comes with collecting surveillance footage of real people, and it keeps legal review short.
Can I get a free sample?
Yes. Most catalogue datasets have a free sample on Kaggle — weapon, knife, fall, drone swarm and more — so you can check the labels and train a baseline before you buy a licence.
How does pricing work?
Catalogue datasets are one-time licences per dataset, research or commercial, for 1,000 images (prices on the pricing page); 5,000 and 10,000 images on request. The False-Positive Stress Test is $1,500. Custom image datasets start at $1,500 and video is quoted per project.
Do you provide data for vision-language models?
Yes. Datasets and custom packs can include natural-language captions and grounded boxes for fine-tuning VLMs, and event-level captions for video clips. See the VLM training data page for formats.
More detail on the use cases, VLM training data and synthetic data for computer vision.
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.