Skip to main content

Data Generation Overview

Overview​

Data generation is one of the key features of DISCOVERSE, including automated data collection and advanced domain randomization techniques. With these tools, users can efficiently generate large amounts of diverse robot learning data, significantly improving model generalization and Sim2Real transfer.

🎯 Core Features​

Automated Data Collection​

  • Multi-robot platform support (AirbotPlay, MMK2, etc.)
  • Parallel data generation for efficiency
  • Synchronized multi-modal data collection (RGB, depth, mask)

Domain Randomization Techniques​

  • Visual transformation based on generative models
  • Optical flow-driven temporal consistency
  • Professional rendering with ComfyUI integration

Data Format Conversion​

  • Support for multiple learning algorithm formats
  • Standard formats such as HDF5, Zarr
  • Automated conversion process

📚 Tutorial List​

Automated Data Collection​

Learn how to use DISCOVERSE's automated data collection system, including:

  • Multi-robot platform data generation
  • Parallel processing configuration
  • Data quality control

Domain Randomization Techniques​

Master advanced domain randomization methods, including:

  • ComfyUI generative model integration
  • Optical flow temporal processing
  • Visual scene transformation