LABELLED TRAINING IMAGES

FOR COMPUTER VISION AI

STOP SPENDING MONTHS GATHERING TRAINING DATA

CREATE CUSTOM DATA FROM REALISTIC 3D SIMULATIONS

STOP SPENDING MONTHS GATHERING TRAINING DATA

CREATE CUSTOM DATA FROM REALISTIC 3D SIMULATIONS

IMAGES

7K

LABELS

79.6K

OBJECTS

2.7K

LABEL TYPES

55

SCENERIOS

6


CURATE

INSTANT DATASETS

Cherry pick existing community images to create new datasets. No waiting, only download what you need
5000 Photos of house interiors from the flo

SEARCH FILTERS

TIME OF DAY
OBJECT TYPE
ENVIRONMENT
SCENARIO
OBJECT COLOUR
RAIN INTENSITY
OBJECT ORIENTATION
SENSOR FOV
SENSOR LOCATION
IMAGE RESOLUTION
CLOUD COVERAGE
OBJECT STATE
OBJECT COLOR
OBJECT DISTANCE

CUSTOMIZABLE

SCENARIOS

Select from a range of customizable 2D & 3D simulation scenarios that match your deployment world conditions

PIXEL PERFECT

LABELS

Move beyond 2D bounding boxes estimates. Every image contains a range of next generation scene and camera information

DON'T JUST TAKE OUR WORD

HOW IS THE INDUSTRY USING SYNTHETIC DATA TODAY

SMART MASK RULES

Avoid cluttering your dataset with irrelevant data. Mix mask rules to ignore objects that may be too small, too far away and more

VISIBILITY

Mask when objects are or arent obstructed

DISTANCE

When objects are a certain distance from the sensor

OBJECT STATE

While objects perform specific actions. e.g. only mask cars that are turned on

ORIENTATION

Mask based on relative angle to the sensor. e.g, only mask when horizontally perpendicular

IMAGE COVERAGE

When objects occupy a specific number of pixels on the image

INSTANCE GROUPED

Mask objects where all instances share a color

INSTANCE

Each object instance gets a unique color mask

PLACEMENT AREA

When objects exist within specific areas. e.g, when only on roads

LOCATION OCCUPANCY

Masked based on occupancy. e.g, mask parking spots with cars

INCLINE

Mask based on incline angle

COMBINED LOCATION

Combine masks of related objects or keep them independent

OBJECT COLOR

When objects have a specific base color

MATCH YOUR EXACT

SENSOR CONFIGURATION

ANY FOV, ANY RESOLUTION, SINGLE SENSOR

STEREOSCOPIC, 360 AND MORE

Any FOV, Any Resolution, Single Sensor, Stereoscopic, 360 and More

COMMUNITY DRIVEN

Browse, download or regenerate community datasets with new rules and settings.

ITERATION MADE EASY

WITH DATASET REPORTS

Access a range of smart reports, graphs and distributions to review datasets and identify trends, high level composition and unaddressed edge cases

DO IT ALL IN YOUR

INTERNET BROWSER

No special hardware required

No downloads required

On your phone or on your desktop

Get started instantly

Get started instantly, no downloads required, no special hardware, on any device

FREQUENTLY ASKED QUESTIONS

The amount of time required can vary depending complexity of a scene and resolution. As a rough broad estimate it takes 5-10 seconds per image once processing begins. Some additional warmup time is present as your custom scene is prepared. Speeds will improve as the website matures
Our physically accurate cameras can be customized with any resolution (capped at 4k), any common fov and can be combined in custom multi sensor arrays
No. Our images are entirely 3d simulated which enables us have full control over every scene. Whether youre generating 10 or 1,000,000 images you will see the same level of quality with no hallucinations
Yes, we offer a range of pricing plans with various features, the cheaper plans only offer public generations while the more business orientated plans offer private generations
Our current focus is the cloud platform. If you'd like to explore an offline option, get in touch. We'll explore if it's viable for your requirements
As we support all resolutions from 100x100 to 4k, web previews are standardized at at a maximum of 800x600 to maintain a smooth website experience. Some compression is also applied to reduce image size which may result in minor artifacts

THE FASTEST WAY TO

BUILD RELIABLE AI

OPEN ANNOTATIONS

All rights reserved.