A very naive compression that only compress on the channel.

Unit Attention class. Todo: mask is not added yet. Parameters ---------- dim: int Input feature dimension. dim_head: int

Unit Attention class. Todo: mask is not added yet. Parameters ---------- dim: int Input feature dimension. dim_head: int

Resnet family to encode image. Parameters ---------- params: dict The parameters of resnet encoder.

A mock class for transform

Data Augmentor. Parameters ---------- augment_config : list A list of augmentation configuration. Attributes ------

A Naive decoder implementation Parameters ---------- params: dict Attributes ---------- num_ch_dec : list The d

Scaled Dot-Product Attention proposed in "Attention Is All You Need" Compute the dot products of the query with all keys, divide each by sqrt

Data rearrange -> swap block -> mlp_head

Vanilla CAV attention.

Scaled Dot-Product Attention proposed in "Attention Is All You Need" Compute the dot products of the query with all keys, divide each by sqrt

Criss-Cross Attention Module reference: https://github.com/speedinghzl/CCNet

Double convoltuion Args: in_channels: input channel num out_channels: output channel num

Double convoltuion Args: in_channels: input channel num out_channels: output channel num

A mock class for Location.

This class is used to manage HD Map. We emulate the style of Lyft dataset. Parameters ---------- world : Carla.World CARLA w

Correct localization error and use Algorithm 1: BBox matching with scores to fuse the proposal BBoxes

Implement the Temporal Encoding (Sinusoid) function.

Implement the Temporal Encoding (Sinusoid) function.

A mock class for Rotation.

Format all scenes in a structured way. Parameters ---------- scenario_params: dict Overall parameters for the replayed scene

Manager for each scene for spawning, moving and destroying. Parameters ---------- folder : str The folder to the current sce

The manager controls all sensor data streaming and dumping for each cav. todo: rsu not considered yet. Parameters ---------- agen