X-gen

Synthetic Data
Generation Solutions

A solution that generates infinite
synthetic images in a virtual environment
and utilizes them for AI training

Overview

X-Gen is a synthetic data generation
solution based on a 3D physics engine.

Leverage the power of 3D physics engines
to train AI models by replacing or
supplementing scarce or non-existent
real-world data.

1 of 2

Main Features

Generate Various Synthetic Data

Generate synthetic data
for different environments.

2D / 3D Simulation

Generate data by adjusting the sun’s position
and illumination throughout the day, as well as
various environmental conditions such as
clouds, fog, snow, rain, shadows, etc.
Conduct virtual simulations of various
situations and events that are difficult to
implement in the real world

Hyper-realistic
Synthetic Image

Design authentic virtual environments using
NVIDIA Omniverse.
Generate higher quality data than
real-world images

Apply Expected Scenario

Run 2D/3D simulations that mimic
real-world patterns

Covering different environments

Create data under various environmental
situations like daytime/nighttime, different
weather conditions, light glare, etc.

Generating virtual environments

Create photorealistic virtual environments

Generate
high-quality
training data

Generate high-quality training data to
improve AI model performance.

Deep Learning

Generate high-quality synthetic data by
utilizing a variety of realistic sensors, including
RGB cameras, IR cameras, LiDAR sensors,
depth sensors, and more.
Complete Instance & Semantic Segmentation
implementation

Dataset Generation

COCO & Pascal VOC format
or customizable datasets
Generate tens to hundreds of thousands of
datasets per day with automatic labeling

Various Camera Sensors

Apply different sensors to improve
performance of synthetic data

Image Segmentation
Detection

Implement seamless instance and semantic segmentation

Create massive datasets

Create diverse and massive datasets with
automatic labeling and 200+ sets per minute
creation speeds

Features &
differentiators

Using X-Gen to develop realistic virtual
data and the NVIDIA Omniverse platform
to distinguish digital twins and different
engineering simulations.

Use Cases

X-GEN

Synthetic Data Generation Solution

Developing a LiDAR-based indoor
driving data generation platform

Generate driving data in indoor
environments using data generated
by photorealistic virtual LiDAR sensors
Create realistic virtual environments
to train AI models

Developing a LiDAR-based on-road
driving data generation platform

Generate driving data in road
environments by leveraging data
generated by realistic virtual LiDAR
sensors
Simulate a wide range of road and
weather conditions

Infrared Synthetic Data Platform
Development

Create synthetic data using data
generated by realistic infrared
camera sensors
Used to train AI models for nighttime
and thermalbased object detection

Developing a synthetic data
generation platform
for small object detection

Deliver high-quality synthetic
datasets for small-scale object
detection
Train AI models in environments
where collecting real-world data
is difficult

Developed a 3D map generation
platform based on satellite maps

Leverage satellite map data to build
photorealistic 3D virtual environments
Train simulations and AI models for
autonomous driving, drones,
and more