Back to portfolio

Synthetic Data

Collecting and labeling real-world 3D data is costly and time-consuming. We build custom synthetic data generation tools and digital twin pipelines that produce large, richly annotated datasets, enabling robust training of machine learning models without extensive physical data capture.

Bin Simulation

We simulate real industrial processes such as parts falling into bins using physics-based simulation. These synthetic scenes closely mirror real-world scanner conditions, including camera parameters, calibration, and light interference, making the generated data directly usable for training AI models.

Read more in our project page Synthetic Data & Digital Twins >
Synthetic render of a bin filled with metallic parts
Our synthetic bin filling pipeline handles everything from rigid-body physical simulation to 3D scanner simulation.

Machine Learning

We use synthetic data to train machine learning models for edge cases in object detection, segmentation, and object classification. The trained networks are used in applications like robotic arm picking, defect detection, or conveyor belt object tracking and counting. We have developed custom training pipelines for PointNet++, YOLO, U-Net, and other models.

See our research on data-driven cardboard box localization >
Machine learning model estimating trained on transparent objects
Machine learning models can estimate the surfaces of transparent objects, which are difficult to capture with 3D scanners.

Rendering and Visualization

We have developed rendering and visualization tools for creating realistic synthetic scenes. We use modern rendering frameworks and engines like Blender, LuxCoreRender, Unreal Engine, and Unity. For photo-realistic simulation of light caustics and transparency effects, we use renderers such as SuperCaustics. We have experience with performance-critical applications in CUDA, Vulkan, and OpenGL.

SMPL meshes with skeletons
Producing realistic images of transparent and translucent objects is computationally challenging due to light propagation and interference. Today, we can use tools like SuperCaustics to simulate these effects in real time.
Synthetic render of play bricks

Interested?

Get in touch with us to choose the right solution for you. Contact us using the information below.

Back to portfolio