davconde / roadperceptionsdg

Road zone Synthetic Data Generator based on Unity Perception.

RoadPerceptionSDG

Road zone Synthetic Data Generator based on Unity Perception

Installation

• Download and install an Unity Editor version satisfying Perception requirements (Unity 2021.3.x for Perception 1.0.0-preview.1).

• Clone this repository:

git clone https://github.com/davconde/RoadPerceptionSDG.git

• Open this project, download the Objects.unitypackage and import it selecting Assets/Import Package/Custom Package... in toolbar.

• Open the Scenes/SDG scene and restart Unity to reload new script references imported from package.

Set Up

• For adding new objects or classes, take as example the subdirectories in Assets/Objects/Foreground. Every Prefab added in the respective directory must be defined as a Prefab Variant with Assets/Object/Foreground/ForegroundObjectdefined as its base.

• Change class objects instantiation probability weights in the Global Settings component of the Game Controller object.

• Randomize lighting parameters in the Directional Light object's Light Randomizer component.

• In the Main Camera object, define pose randomization with the Camera Randomizer component, and general annotation settings in the Perception Camera component.

Use

In the Perception Camera component of Main Camera, check the labelers to use. Data will be stored in %userprofile%\AppData\LocalLow\DefaultCompany\SDG. Click on Play button to start the simulation.

For dataset processing and annotation conversion, use pysolotools.