Edge Residual Block. Args: in_channels (int): The input channels of this module. out_channels (int): The output channels of this

Class for sampling specific category of ground truths. Args: sample_list (list[dict]): List of samples. name (str | None): The ca

ModuleDict in openmmlab. Args: modules (dict, optional): a mapping (dictionary) of (string: module) or an iterable of key-val

Image to Patch Embedding V2. We use a conv layer to implement PatchEmbed. Args: in_channels (int): The num of input channels. Default

Merge patch feature map. This layer use nn.Unfold to group feature map by kernel_size, and use norm and linear layer to embed grouped feature

Squeeze-and-Excitation Module. Args: in_channels (int): The input (and output) in_channels of the SE layer. squeeze_channels (Non

Shift Window Multihead Self-Attention Module. Args: embed_dims (int): Number of input channels. num_heads (int): Number of attent

Spatial Transform. NOTE: The output feature map has the same size with the input. Args: conv_group_cfg: Conv group in localizat

Args: embed_dims (int): The feature dimension. num_heads (int): Parallel attention heads. feedforward_channels (int): The

Implements one stage in Swin Transformer. Args: embed_dims (int): The feature dimension. num_heads (int): Parallel attention head

Window based multi-head self-attention (W-MSA) module with relative position bias. Args: embed_dims (int): Number of input channels.

Fuse multi-modality features based on attention mechanism Args: attention_only (bool): Whether use attention only. If true, there will

BEVFusion. Args: encoders (list): List of encoders' config. fuser (dict): Config of fuser. decoder (dict): Config of deco

Base class for fusion_models.

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

Convolution Backbone Map Encoder. Args: conv_group_cfg: (dict): The config of convolution group. - in_channels (int): The in

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

Gather multi edge residual blocks as a group.

r"""Feature Pyramid Network. This is an implementation of paper `Feature Pyramid Networks for Object Detection <https://arxiv.org/abs/1612.03

Focus width and height information into channel space. Args: in_channels (int): The input channels of this Module. out_channels (

L2 loss. Args: reduction (str, optional): The method to reduce the loss. Options are "none", "mean" and "sum". loss_w

Load BEV map mask Args: data_root (str): The path of the dateset root directory. xbound (Tuple(float, float, float)): xmin, xmax,

Load multi channel images from a list of separate channel files. Expects results['image_paths'] to be a list of filenames. Args: to_

Class for sampling data from the ground truth database. Compared to `DataBaseSampler`, `MapEnhancedDataBaseSampler` consider surround

r"""NuScenes Dataset. This class serves as the API for experiments on the NuScenes Dataset. Please refer to `NuScenes Dataset <https://www.n

Args: - stop_epoch (int | None): The number of stop epoch. If None, will not stop.

Resize map size directly by interpolation.

Map Encoder with spatial transformer. Args: output_size: Output size of the interpolation when resizing feature map. ONLY wh

Swin Transformer A PyTorch implement of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows` - https://arxiv.