virtual-camera 1.0.3

Virtual Camera

Virtual Camera

Install

pip install virtual_camera

Instructions

Coordinate systems

STYLE	X-Y-Z
MOTOVIS	right-forward-up
openGL	right-up-backward
camera	right-down-forward
pytorch3d	left-up-forward

Create virtual cameras

Following are default virtual camera setups.

VCAMERA_PERSPECTIVE_FRONT = create_virtual_perspective_camera((1280, 960), (-90, 0, 0), (0, 1.5, 1.5))
VCAMERA_PERSPECTIVE_FRONT_LEFT = create_virtual_perspective_camera((1280, 960), (-90, 0, 45), (-1, 2, 1))
VCAMERA_PERSPECTIVE_FRONT_RIGHT = create_virtual_perspective_camera((1280, 960), (-90, 0, -45), (1, 2, 1))
VCAMERA_PERSPECTIVE_BACK = create_virtual_perspective_camera((1280, 960), (-90, 0, 180), (0, -1, 1))
VCAMERA_PERSPECTIVE_BACK_LEFT = create_virtual_perspective_camera((1280, 960), (-90, 0, 135), (-1, 2, 1))
VCAMERA_PERSPECTIVE_BACK_RIGHT = create_virtual_perspective_camera((1280, 960), (-90, 0, -135), (1, 2, 1))

VCAMERA_FISHEYE_FRONT = create_virtual_fisheye_camera((1024, 640), (-120, 0, 0), (0, 3.5, 0.5))
VCAMERA_FISHEYE_LEFT = create_virtual_fisheye_camera((1024, 640), (-135, 0, 90), (-1, 2, 1))
VCAMERA_FISHEYE_RIGHT = create_virtual_fisheye_camera((1024, 640), (-135, 0, -90), (1, 2, 1))
VCAMERA_FISHEYE_BACK = create_virtual_fisheye_camera((1024, 640), (-120, 0, 180), (0, -1, 0.5))

Convert from a real image

import yaml
import virtual_camera as vc

1. Initialize the a camera from a real image and its calibration file.

yaml_file = 'data/calibration.yml'
cfg = yaml.safe_load(open(yaml_file, 'r'))
camera_real = vc.FisheyeCamera.init_from_motovis_cfg(cfg)

2. Read the real image.

src_image = plt.imread('data/src_image.jpg')

3. Convert the image to a virtual image. Alongside with mask.

dst_image, mask = vc.render_image(src_image, camera_real, vc.VCAMERA_PERSPECTIVE_FRONT)