Skip to main content
Augment generates high-fidelity variations of RGB data to improve the robustness of downstream models. It supports adding realism, weather, and lighting shifts to real or synthetic datasets while maintaining ground-truth integrity.

​
Overview

The Augment pipeline features task-specific workflows and presets designed for high-fidelity data synthesis. You can test the augmentation parameters on individual images to verify the outputs and perform automated batch runs after finalizing the augmentation parameters. The augmented data is available for immediate download or can be saved to your library for future use.

​
Workflows

​
1. Sim-to-Real

The Sim-to-Real workflow is designed to add photorealism to synthetic datasets (such as simulation images). Built-in and custom scene presets provide options for diverse lighting and weather conditions (Daylight Clear, Daylight Fog, Daylight Rain, Daylight Snow, Night/Dusk Clear, Night/Dusk Fog, Night/Dusk Rain, Night/Dusk Snow).
Sim-to-Real workflows require synthetic datasets containing rgb_raw and depth_carla modalities.

​
2. Real-to-Real

The Real-to-Real workflow augments existing real or synthetic RGB datasets across lighting and weather conditions while preserving the exact scene composition, vehicles, buildings, and objects.