Generate hard-to-capture data in a virtual environment
X-GEN is a synthetic data solution that recreates environments that are difficult to film or collect enough data from, then generates the images and sensor data needed to train AI models.
X-GEN key capabilities
Synthetic data generation
Virtual environment and scenario setup
Multi-sensor data generation
Automatic labeling
X-GEN PROCESS
From scenes built for your training goal to AI-ready training data
STEP 01
Build the virtual scene
Set up spaces, objects, cameras and motion
STEP 02
Set environment and scenario
Configure weather, time of day, lighting and object placement
STEP 03
Generate data and labels
Produce images, sensor data and training labels
STEP 04
Train and validate AI
Use alongside real data to train and validate models
Multiple training datasets from a single scene
One virtual scene yields photorealistic images along with segmentation, normal and depth data, giving you training data in several forms at once.
VIRTUAL ENVIRONMENT & SYNTHETIC DATA
Recreate environments and situations that are hard to film in real life
Indoor spaces, roads and industrial sites are rebuilt in 2D and 3D. By changing time of day, weather, lighting, objects and camera conditions, you can turn scenes that are impractical to shoot repeatedly into data.
Recreates the spaces and objects your models learn from in 2D and 3D
Generates high-quality synthetic data that reflects real-world materials, light and shadow
Reproduces situations and events that can occur in the real world inside a virtual environment
Applies day and night, rain, snow, fog, lighting and other environmental conditions
Builds lifelike spaces and the 3D objects your training needs, such as vehicles and equipment
SENSOR DATA & AUTOMATIC LABELING
Generate diverse sensor data and training labels together
X-GEN produces RGB, IR, LiDAR and depth data together with training labels, and can package them as COCO, Pascal VOC or custom datasets.
Sensor data
- RGB
- IR
- LiDAR
- Depth
- Normal
Label data
- 2D and 3D bounding box
- Semantic segmentation
- Instance segmentation
Supported formats
- COCO
- Pascal VOC
- Custom dataset
APPLICATION AREAS
Generate the training data you need where real data is hard to collect
With X-GEN you can generate synthetic data that fits your training goal even in fields where filming is difficult or where varied sensors and environmental conditions are required.
Indoor navigation data
Builds indoor spaces and obstacles virtually, then uses LiDAR sensors to generate movement paths and surrounding spatial data
Road driving data
Recreates roads, vehicles and surroundings to generate LiDAR data across a range of driving conditions
Infrared synthetic data
Generates infrared data for conditions such as night or low light, where ordinary cameras cannot collect enough samples
Satellite-based 3D maps
Builds wide-area 3D environments from satellite maps and spatial data for autonomous driving and drone simulation
Small object detection
Recreates small objects that are hard to capture in sufficient quantity in the real world
X-GEN generates synthetic data using virtual environments and simulation technology built on NVIDIA Omniverse™.
Check our frequently asked questions
VIEW ALLCan we use the 3D data we already have?
Yes. Existing 3D data or site data can be used to build the virtual environment. Which data can be used and how it is built depend on the form of your data and the purpose; where needed, we can construct a 3D virtual environment closely resembling the real site for simulation and AI training.
We want a digital twin built for our industry and site.
Xiilab proposes digital twin approaches suited to a range of industries and purposes, including manufacturing, construction, and defense. The simulation environment and features can be configured to match your purpose and site conditions.
How does X-GEN use NVIDIA Omniverse?
X-GEN generates synthetic data using virtual environments and simulation technology built on NVIDIA Omniverse. It recreates real environments in virtual space and generates data under varied conditions for AI training and validation.
Can we validate in a virtual environment before building the real site?
With a digital twin, you can review a range of conditions and scenarios in a virtual environment before building. Depending on your purpose, we can set up simulation, virtual validation, or AI training environments.
We would like advice on how to apply a digital twin before adopting one.
Once we confirm your goals and site environment, we will outline the scope of work required and how it can be applied.