Road zone Synthetic Data Generator based on Unity Perception
-
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.
-
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 withAssets/Object/Foreground/ForegroundObject
defined as its base. -
Change class objects instantiation probability weights in the
Global Settings
component of theGame Controller
object. -
Randomize lighting parameters in the
Directional Light
object'sLight Randomizer
component. -
In the
Main Camera
object, define pose randomization with theCamera Randomizer
component, and general annotation settings in thePerception Camera
component.
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.