Processors for keypoints

ChangeKeypointsCoordinateSystem

paz.processors.keypoints.ChangeKeypointsCoordinateSystem()

Changes keypoints 2D coordinate system using box2D coordinates to locate the new origin at the openCV image origin (top-left).

[source]

DenormalizeKeypoints

paz.processors.keypoints.DenormalizeKeypoints()

Transform normalized keypoints coordinates into image-size coordinates.

Arguments

image_size: List of two floats having height and width of image.

[source]

NormalizeKeypoints

paz.processors.keypoints.NormalizeKeypoints(image_size)

Transform keypoints in image-size coordinates to normalized coordinates.

Arguments

image_size: List of two ints indicating ''(height, width)''

[source]

PartitionKeypoints

paz.processors.keypoints.PartitionKeypoints()

Partitions keypoints from shape [num_keypoints, 2] into a list of the form ((2), (2), ....) and length equal to num_of_keypoints.

[source]

ProjectKeypoints

paz.processors.keypoints.ProjectKeypoints(projector, keypoints)

Projects homogenous keypoints (4D) in the camera coordinates system into image coordinates using a projective transformation.

Arguments

projector: Instance of ''paz.models.Project''.
keypoints: Numpy array of shape ''(num_keypoints, 3)''

[source]

RemoveKeypointsDepth

paz.processors.keypoints.RemoveKeypointsDepth()

Removes Z component from keypoints.

[source]

TranslateKeypoints

paz.processors.keypoints.TranslateKeypoints()

Applies a translation to keypoints. The translation is a list of length two indicating the x, y values.

[source]

DenormalizeKeypoints2D

paz.processors.keypoints.DenormalizeKeypoints2D()

Transform normalized keypoints coordinates into image-size coordinates.

Arguments

image_size: List of two floats having height and width of image.

[source]

NormalizeKeypoints2D

paz.processors.keypoints.NormalizeKeypoints2D(image_size)

Transform keypoints in image-size coordinates to normalized coordinates.

Arguments

image_size: List of two ints indicating ''(height, width)''

[source]

ArgumentsToImageKeypoints2D

paz.processors.keypoints.ArgumentsToImageKeypoints2D()

Convert array arguments into UV coordinates.

Image plane

(0,0)--------> (U) | | | v

(V)

Arguments

row_args: Array (num_rows).
col_args: Array (num_cols).

Returns

Array (num_cols, num_rows) representing points2D in UV space.

Notes

Arguments are row args (V) and col args (U). Image points are in UV coordinates; thus, we concatenate them in that order i.e. [col_args, row_args]

[source]

ScaleKeypoints

paz.processors.keypoints.ScaleKeypoints(scale=1, shape=(128, 128))

Scale keypoints to input image shape.

Arguments

keypoints: Array. Detected keypoints by the model
image: Array. Input image.

Returns

Scaled keypoints: Array. keypoints scaled to input image shape.

[source]

ComputeOrientationVector

paz.processors.keypoints.ComputeOrientationVector(parents)

Calculate the orientation of keypoints links with 3D keypoints.

Arguments

keypoints: Array. 3D keypoints

Returns

orientation: Array. Orientation of keypoint links

[source]

MergeKeypoints2D

paz.processors.keypoints.MergeKeypoints2D(args_to_mean)

[source]

FilterKeypoints2D

paz.processors.keypoints.FilterKeypoints2D(args_to_mean, h36m_to_coco_joints2D)

[source]

StandardizeKeypoints2D

paz.processors.keypoints.StandardizeKeypoints2D(data_mean2D, data_stdev2D)

[source]

DestandardizeKeypoints2D

paz.processors.keypoints.DestandardizeKeypoints2D(data_mean3D, data_stdev3D, dim_to_use)

[source]

OptimizeHumanPose3D

paz.processors.keypoints.OptimizeHumanPose3D(args_to_joints3D, solver, camera_intrinsics)

Optimize human 3D pose

Arguments

solver: library solver camera_intrinsics: camera intrinsic parameters

Returns

keypoints3D, optimized keypoints3D