Camera Class Reference¶

class MultiViewTracks.Camera(id, name, fisheye, extrinsics, intrinsics, tracks, verbose=True)¶

This is a class containing per camera preprocessing methods for triangulating tracks using the Scene class.

id¶

The camera id within a COLMAP reconstruction

Type:	int

name¶

A common prefix used for all images of this camera in a COLMAP reconstruction

Type:	str

fisheye¶

Was OPENCV_FISHEYE used as COLMAP camera model?

Type:	bool

image_names¶

Contains the image names of the reconstructed views of this camera

Type:	np.ndarray

view_idx¶

Contains the frame indices of all reconstructed views of this camera

Type:	np.ndarray

n_views¶

Number of reconstrueced views of this camera

Type:	int

R¶

A Rotation instance holding extrinsic parameters (rotations) for this camera

Type:	scipy.spatial.transform.Rotation

r¶

Stores the rotation matrices retrieved from R

Type:	np.ndarray

t¶

Stores COLMAP extrinsic camera parameters for each view of this camera

Type:	np.ndarray

k¶

The camera matrix of this camera

Type:	np.ndarray

d¶

The distortion parameters of this camera

Type:	np.ndarray

tracks¶

Contains the tracks visible from this camera or None

Type:	dict

tracks_undistorted¶

Contains undistorted tracks from Camera.undistort_tracks or None

Type:	dict

tracks_projected¶

Contains transformed tracks from Camera.project_tracks or None

Type:	dict

tracks_reprojected¶

Contains reprojected tracks from Camera.reproject_tracks or None

Type:	dict

verbose¶

Do you want some verbosity? Defaults to true

Type:	bool

frames_in_view(i)¶: Returns the frame indices in which individual i is observed in the camera views.

get_rotations()¶: Returns a np.ndarray of rotation matrices transformed from R.

interpolate()¶: Interpolates the camera path by linearly interpolating t and using SLERP for R.

position(idx, i, kind='')¶

Returns the position of individual i at the specified frame index.

Parameters:	idx (int) – The frame index i (int) – The individual’s identity kind (str, optional) – One of “”, “undistorted”, “projected”. Defaults to “”
Returns:	The position of individual i at frame index idx
Return type:	np.ndarray

project_tracks()¶: Projects the tracks to world coordinates with unknown depth using r and t, Camera.undistort_tracks first if necessary.

projection_center(idx)¶: Return the 3d coordinates of the idx-th view projection center.

reproject_tracks(tracks_3d)¶

Projects given tracks from world coordinates to image coordinates using r and t, and distortion model. The tracks must be in the original coordinate system, i.e. they should not be rotated or scaled.

Parameters:	tracks_3d (dict) – The 3D tracks in the world coordiate system to be reprojected.

undistort_tracks()¶: Undistorts the tracks in image coordinates to normalized coordinates.

view(idx)¶: Returns the projection matrix of the camera at the specified frame index.