Cameras

For both aerial and wheeled robots, AirGen offers a comprehensive suite of camera sensors that play a crucial role in simulation and data generation. These sensors provide various modalities representative of commonly used sensor types and are essential for creating high-fidelity data for AI model training or testing.

Camera Modalities

AirGen has several classes of camera modalities, offering unique data essential for different aspects of robotic perception and navigation. Below is a list of modalities that are available.

  • RGB: The most commonly used camera sensor: captures RGB/scene data.

  • Depth: Capture depth information to understand the spatial structure of the environment.

  • Segmentation: Provide segmented views of the scene, useful for object recognition and scene understanding.

  • Infrared: Capture thermal data, enhancing vision in low-light conditions.

  • Surface Normals: Provide data on surface orientation, critical for detailed object understanding.

  • Optical Flow: Track motion across frames, aiding in dynamic scene analysis.

Available Camera Modalities

Each camera modality is identified by an ImageType integer. The table below lists the current camera sensor modalities available in AirGen:

Camera Type

ImageType

Scene (RGB)

0

DepthPlanar

1

DepthPerspective

2

DepthVis

3

DisparityNormalized

4

Segmentation

5

SurfaceNormals

6

Infrared

7

OpticalFlow

8

OpticalFlowVis

9

Camera Placement and Configuration

AirGen vehicles typically feature five strategically placed cameras: front_center, front_right, front_left, bottom_center, and back_center. These positions ensure comprehensive coverage for simulation scenarios. The desired camera can be targeted by specifically stating the string from the ones listed above.

Image Capture

Further Reading

For more detailed information on each camera modality and their specific use cases, refer to the following sections: