Compose multiple transforms sequentially. The pipeline registry of mmdet3d separates with mmdet, however, sometimes we may need to use mmdet's

3D boxes of instances in LIDAR coordinates. Coordinates in LiDAR: .. code-block:: none up z x fron

3D boxes of instances in Depth coordinates. Coordinates in Depth: .. code-block:: none up z y front (yaw=-0.5*

3D boxes of instances in CAM coordinates. Coordinates in camera: .. code-block:: none z front (yaw=-0.5*pi)

3d residual block for ImVoxelNeck. Args: in_channels (int): Number of channels in input tensor. out_channels (int): Number of

r"""More general bbox head, with shared conv layers and two optional separated branches. .. code-block:: none /->

Very simple multi-layer perceptron (also called FFN)

Point feature propagation module used in PointNets. Propagate the features from one set to another. Args: mlp_channels (list[in

ScanNet data. Generate scannet infos for scannet_converter. Args: root_path (str): Root path of the raw data. split (s

Tree in DLANet. Args: levels (int): The level of the tree. block (nn.Module): The block module in tree. in_channels:

Edge Fusion Module for feature map. Args: out_channels (int): The number of output channels. feat_channels (int): The number

Iterative Deep Aggregation (IDA) Upsampling module to upsample features of different scales to a similar scale. Args: out_channel

KITTI predictions to Waymo converter. This class serves as the converter to change predictions from KITTI to Waymo format. Args:

SUNRGBD data. Generate scannet infos for sunrgbd_converter. Args: root_path (str): Root path of the raw data. split (s

ScanNet dataset used to generate infos for semantic segmentation task. Args: data_root (str): Root path of the raw data. ann_

Sparse basic block for PartA^2. Sparse basic block implemented with submanifold sparse convolution. Args: inplanes (int): inpla

r"""Online visualizer implemented with Open3d. Args: points (numpy.array, shape=[N, 3+C]): Points to visualize. The Points

Vote module. Generate votes from seed point features. Args: in_channels (int): Number of channels of seed point features.

Voxel generator in numpy implementation. Args: voxel_size (list[float]): Size of a single voxel point_cloud_range (list[float

Parallel version of waymo dataset information gathering. Waymo annotation format version like KITTI: { [optional]points: [N,

Axis-aligned 3D Overlaps (IoU) Calculator.

Class for sampling specific category of ground truths. Args: sample_list (list[dict]): List of samples. name (str, optional):

A wrapper of class sampled dataset with ann_file path. Implementation of paper `Class-balanced Grouping and Sampling for Point Cloud 3D Object

Point feature aggregation module used in DGCNN. Aggregate all the features of points. Args: mlp_channels (list[int]): List of m

Point feature propagation module used in DGCNN. Propagate the features from one set to another. Args: mlp_channels (list[int]):

Point graph feature module used in DGCNN. Args: mlp_channels (list[int]): Specify of the dgcnn before the global pooling

Deep Layer Aggregation (DLA) Upsampling module for different scales feature extraction, upsampling and fusion, It consists of groups of IDAu

Given the raw data, generate the ground truth database. This is the parallel version. For serialized version, please refer to `create_ground

Sampling Points. Sampling points with given index.

Single instance for ScanNet evaluator. Args: mesh_vert_instances (np.array): Instance ids for each point. instance_id: Id of

A simple MLP module. Pass features (B, C, N) through an MLP. Args: in_channels (int, optional): Number of channels of input fea

Non-CUDA version of PAConv. PAConv stores a trainable weight bank containing several kernel weights. Given input points and features, it c

CUDA version of PAConv that implements a cuda op to efficiently perform kernel assembling. Different from vanilla PAConv, the input featur

Pillar Feature Net Layer. The Pillar Feature Net is composed of a series of these layers, but the PointPillars paper results only used a s

object candidate point prediction from seed point features. Args: in_channel (int): number of channels of seed point features.

Class for single ScanNet RGB-D image processing.

Points of instances in LIDAR coordinates. Args: tensor (torch.Tensor | np.ndarray | list): a N x points_dim matrix. points_di

Root in DLANet. Args: in_channels (int): Input feature channel. out_channels (int): Output feature channel. norm_cfg

S3DIS data. Generate s3dis infos for s3dis_converter. Args: root_path (str): Root path of the raw data. split (str, op

S3DIS dataset used to generate infos for semantic segmentation task. Args: data_root (str): Root path of the raw data. ann_fi

r"""ScoreNet that outputs coefficient scores to assemble kernel weights in the weight bank according to the relative position of point pairs.

Class for single ScanNet scene processing. Single scene file contains multiple RGB-D images.

SeparateHead for CenterHead. Args: in_channels (int): Input channels for conv_layer. heads (dict): Conv information.

Voxel Feature Encoder layer. The voxel encoder is composed of a series of these layers. This module do not support average pooling and onl

Calculate the number of points inside the ground truth box. This is the parallel version. For the serialized version, please refer to `_calc

r"""S3DIS Dataset for Semantic Segmentation Task. This class is the inner dataset for S3DIS. Since S3DIS has 6 areas, we often train on 5

Get the affine transform matrices to the target size. Different from :class:`RandomAffine` in MMDetection, this class can calculate the af

Aligned 3D Anchor Generator by range. This anchor generator uses a different manner to generate the positions of anchors' centers from :cl

3D Anchor Generator by range for per class. This anchor generator generates anchors by the given range for per class. Note that feature ma

Anchor head for SECOND/PointPillars/MVXNet/PartA2. Args: num_classes (int): Number of classes. in_channels (int): Number of c

3D Anchor Generator by range. This anchor generator generates anchors by the given range in different feature levels. Due the convent

Anchor free bbox coder for 3D boxes. Args: num_dir_bins (int): Number of bins to encode direction angle. with_rot (bool): Whe

Anchor-free head for monocular 3D object detection. Args: num_classes (int): Number of categories excluding the background

Mixin class for target assigning of dense heads.

Calculate the IoU loss (1-IoU) of axis aligned bounding boxes. Args: reduction (str): Method to reduce losses. The valid

BBox3DL1Cost. Args: weight (int | float, optional): loss_weight

Filter background points near the bounding box. Args: bbox_enlarge_range (tuple[float], float): Bbox enlarge range.

Base class for BaseDecodeHead. Args: channels (int): Channels after modules, before conv_seg. num_classes (int): Number of cl

Base class for detectors.

Base class for 3d RoIHeads.

Base class for 3D segmentors. The main difference with `BaseSegmentor` is that we modify the keys in data_dict and use a 3D seg specific v

Base module for point graph feature module used in DGCNN. Args: radii (list[float]): List of radius in each knn or ball query.

Base class for 3D Boxes. Note: The box is bottom centered, i.e. the relative position of origin in the box is (0.5, 0.5, 0).

Base class for Monocular 3D DenseHeads.

Base class for PointNet.

Base module for point set abstraction module used in PointNets. Args: num_point (int): Number of points. radii (list[float]):

Base class for Points. Args: tensor (torch.Tensor | np.ndarray | list): a N x points_dim matrix. points_dim (int, optional):

Base Shape-aware Head in Shape Signature Network. Note: This base shape-aware grouping head uses default settings for small o

BasicBlock in DLANet. Args: in_channels (int): Input feature channel. out_channels (int): Output feature channel. no

3D IoU Calculator. Args: coordinate (str): The coordinate system, valid options are 'camera', 'lidar', and 'depth'.

Nearest 3D IoU Calculator. Note: This IoU calculator first finds the nearest 2D boxes in bird eye view (BEV), and then calcul

r"""Enum of different ways to represent a box. Coordinates in LiDAR: .. code-block:: none up z

Points of instances in CAM coordinates. Args: tensor (torch.Tensor | np.ndarray | list): a N x points_dim matrix. points_dim

CenterHead for CenterPoint. Args: in_channels (list[int] | int, optional): Channels of the input feature map. Default: [1

Base class of Multi-modality VoxelNet.

Bbox coder for CenterPoint. Args: pc_range (list[float]): Range of point cloud. out_size_factor (int): Downsample factor of t

Calculate Chamfer Distance of two sets. Args: mode (str): Criterion mode to calculate distance. The valid modes are smoot

Collect data from the loader relevant to the specific task. This is usually the last stage of the data loader pipeline. Typically keys is

Absolute position embedding with Conv learning. Args: input_channel (int): input features dim. num_pos_feats (int, optional):

r"""Enum of different ways to represent a box and point cloud. Coordinates in LiDAR: .. code-block:: none

Customized 3D dataset. This is the base dataset of SUNRGB-D, ScanNet, nuScenes, and KITTI dataset. .. code-block:: none [

Customized 3D dataset for semantic segmentation task. This is the base dataset of ScanNet and S3DIS dataset. Args: data_root (s

r"""DCNSeparateHead for CenterHead. .. code-block:: none /-----> DCN for heatmap task -----> heatmap task. feature

Backbone network for DGCNN. Args: in_channels (int): Input channels of point cloud. num_samples (tuple[int], optional): The n

r"""DGCNN decoder head. Decoder head used in `DGCNN <https://arxiv.org/abs/1801.07829>`_. Refer to the `reimplementation code <https:

DLA Neck. Args: in_channels (list[int], optional): List of input channels of multi-scale feature map. start_leve

r"""`DLA backbone <https://arxiv.org/abs/1707.06484>`_. Args: depth (int): Depth of DLA. Default: 34. in_channels (int, optio

Class for sampling data from the ground truth database. Args: info_path (str): Path of groundtruth database info. data_root (

Default formatting bundle. It simplifies the pipeline of formatting common fields, including "img", "proposals", "gt_bboxes", "gt_labels",

Default formatting bundle. It simplifies the pipeline of formatting common fields for voxels, including "proposals", "gt_bboxes", "gt_labe

Bbox Coder for 3D boxes. Args: code_size (int): The dimension of boxes to be encoded.