torch.nn.init.kaiming_uniform_()' with 'groups'. If 'mode=="fan_out"', fan is divided by 'groups', yielding larger std of weights.

If distributed, only the master executes the function. Usage: @master_only def foo(a, b): ...

Returns a 4x4 homogeneous matrix of the form [R t; 0 1] Returns ---------- result: np.ndarray 4x4 homogeneous mat

Process the pair of inputs and outputs. If they contain batches, the pairs can be consumed one-by-one using `zip`: .. code-b

Process the pair of inputs and outputs. If they contain batches, the pairs can be consumed one-by-one using `zip`: .. code-b

Render radar point cloud in BEV perspective. Parameters ---------- radar_point_cloud: numpy array with shape (N, 3)

Preparation for a new round of evaluation. Should be called before starting a round of evaluation.

Preparation for a new round of evaluation. Should be called before starting a round of evaluation.

Adapted from: https://github.com/TRI-ML/packnet-sfm_internal/blob/919ab604ae2319e4554d3b588877acfddf877f9c/packnet_sfm/datasets/augmentat

Return the rotation component of the pose as a Quaternion object. Returns ---------- self.quat: Quaternion Rotati

Returns the 3x3 rotation matrix (R) Returns ---------- result: np.ndarray 3x3 rotation matrix

Return the translation component of the pose as a np.ndarray. Returns ---------- self.tvec: np.ndarray Translatio

A batch collator that does nothing.

Parameters ---------- x: Dict One 'dataset_dict'. Returns ------- viz_images: Dict[np.a