Return a copy of this pose object. Returns ---------- result: Pose Copied pose object.

Returns a new Pose that corresponds to the inverse of this one. Returns ---------- result: Pose Inverted

This function is taken from the original tf repo. It ensures that all layers have a channel number that is divisible by 8 It can be seen here

Repeat `val` for `repeat_time` times and return the list or val if list/tuple.

Parameters ---------- vecs: Iterable[np.ndarray] Iterable of 10D pose representation. intrinsics: np.nda

configurable is experimental. out_features (list[str]): name of the layers whose outputs should be returned in f

Args: tensors: a tuple or list of `torch.Tensor`, each of shape (Hi, Wi) or (C_1, ..., C_K, Hi, Wi) where K >= 1.

Args: input_ch(int) : the number of input channel out_features (list[str]): name of the layers whose outputs should

Initialize pose from 4x4 transformation matrix Parameters ---------- transformation_matrix: np.ndarray 4x4 contai

Parameters ---------- instances: List[Instances] Predicted instances. group_idxs: dict Mapping from nuScene's `sampl

Adapted from AdelaiDet: https://github.com/aim-uofa/AdelaiDet/blob/master/adet/utils/comm.py

Sync a directory from source to target (either local to s3, s3 to s3, s3 to local) Parameters ---------- source: str Directo

Parameters ---------- boxes3d: Boxes3D 3D boxes in camera frame. scores: Tensor 1D score vector. Must be of same si

Parameters ---------- cfg: CfgNode [description] input_shape: ShapeSpec [description]

Parameters ---------- img: np.ndarray RGB image. (H, W, 3) boxes3d: tridet.structures.Boxes3D or np.ndarray If np.nd

Almost identical to detectron2.evaluation.testing:flatten_result_dict()', but using 'OrderedDict' -------------------------------------------

Creates a mosaic from list of images. Parameters ---------- items: list of np.ndarray List of images to mosaic. scale: float

Smooth L1 loss defined in the Fast R-CNN paper as: | 0.5 * x ** 2 / beta if abs(x) < beta smoothl1(x) = |

Parameters ---------- points2d: Tensor xy coordinates. shape=(N, ..., 2) E.g., (N, 2) or (N, K, 2) or (N, H, W, 2) i

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

This function is used to transpose image first training targets to level first ones :return: level first training targets

Build a backbone from `cfg.FE.BUILDER` Returns: an instance of :class:`Backbone`

Create a VoVNet instance from config. Returns: VoVNet: a :class:`VoVNet` instance.

Build D2 dataset (i.e. List[Dict]), given a dataset implementing recipe for building its item. This is useful when a __getitem__() takes much tim

1x1 convolution with padding

3x3 convolution with padding

Convert a single 3D bounding box (GenericBoxes3D) to KITTI convention. i.e. for evaluation. We assume the box is in the reference frame of camera_

Get mAP from precision. Parameters ---------- precision: np.ndarray Numpy array of precision curves at different

Shard number into evenly sized parts. `Remaining` values are put into the last shard. Parameters ---------- num: int

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

rotated box iou running in gpu. 500x faster than cpu version (take 5ms in one example with numba.cuda code). convert from [this project](

Copy single file from local to s3, s3 to local, or s3 to s3. Parameters ---------- source_path: str Path of file to copy tar

An equivalent of `torch.as_tensor`, but works under tracing if input is a list of tensor. `torch.as_tensor` will record a constant in tracing

Execute inference on a list of inputs, using batch size = self.batch_size, instead of the length of the list. Inputs & outpu

Execute inference on a list of inputs, using batch size = self.batch_size, instead of the length of the list. Inputs & outpu

Parameters ---------- box1, box2: (x1, y1, x2, y2)

Convert KITTI annotation data frame to 3D bounding box annotations. Labels are provided in the reference frame of camera_2. NOTE: Anno

Velocity of ego-vehicle in m/s.