navsim/agents/transfuser/transfuser_model.py

from typing import Dict
import numpy as np
import torch
import torch.nn as nn

from navsim.agents.transfuser.transfuser_config import TransfuserConfig
from navsim.agents.transfuser.transfuser_backbone import TransfuserBackbone
from navsim.common.enums import StateSE2Index
from navsim.agents.transfuser.transfuser_features import BoundingBox2DIndex


class TransfuserModel(nn.Module):
    def __init__(self, config: TransfuserConfig):

        super().__init__()

        self._query_splits = [
            1,
            config.num_bounding_boxes,
        ]

        self._config = config
        self._backbone = TransfuserBackbone(config)

        self._keyval_embedding = nn.Embedding(
            8**2 + 1, config.tf_d_model
        )  # 8x8 feature grid + trajectory
        self._query_embedding = nn.Embedding(sum(self._query_splits), config.tf_d_model)

        # usually, the BEV features are variable in size.
        self._bev_downscale = nn.Conv2d(512, config.tf_d_model, kernel_size=1)
        self._status_encoding = nn.Linear(4 + 2 + 2, config.tf_d_model)

        self._bev_semantic_head = nn.Sequential(
            nn.Conv2d(
                config.bev_features_channels,
                config.bev_features_channels,
                kernel_size=(3, 3),
                stride=1,
                padding=(1, 1),
                bias=True,
            ),
            nn.ReLU(inplace=True),
            nn.Conv2d(
                config.bev_features_channels,
                config.num_bev_classes,
                kernel_size=(1, 1),
                stride=1,
                padding=0,
                bias=True,
            ),
            nn.Upsample(
                size=(config.lidar_resolution_height // 2, config.lidar_resolution_width),
                mode="bilinear",
                align_corners=False,
            ),
        )

        tf_decoder_layer = nn.TransformerDecoderLayer(
            d_model=config.tf_d_model,
            nhead=config.tf_num_head,
            dim_feedforward=config.tf_d_ffn,
            dropout=config.tf_dropout,
            batch_first=True,
        )

        self._tf_decoder = nn.TransformerDecoder(tf_decoder_layer, config.tf_num_layers)
        self._agent_head = AgentHead(
            num_agents=config.num_bounding_boxes,
            d_ffn=config.tf_d_ffn,
            d_model=config.tf_d_model,
        )

        self._trajectory_head = TrajectoryHead(
            num_poses=config.trajectory_sampling.num_poses,
            d_ffn=config.tf_d_ffn,
            d_model=config.tf_d_model,
        )

    def forward(self, features: Dict[str, torch.Tensor]) -> Dict[str, torch.Tensor]:

        camera_feature: torch.Tensor = features["camera_feature"]
        lidar_feature: torch.Tensor = features["lidar_feature"]
        status_feature: torch.Tensor = features["status_feature"]

        batch_size = status_feature.shape[0]

        bev_feature_upscale, bev_feature, _ = self._backbone(camera_feature, lidar_feature)

        bev_feature = self._bev_downscale(bev_feature).flatten(-2, -1)
        bev_feature = bev_feature.permute(0, 2, 1)
        status_encoding = self._status_encoding(status_feature)

        keyval = torch.concatenate([bev_feature, status_encoding[:, None]], dim=1)
        keyval += self._keyval_embedding.weight[None, ...]

        query = self._query_embedding.weight[None, ...].repeat(batch_size, 1, 1)
        query_out = self._tf_decoder(query, keyval)

        bev_semantic_map = self._bev_semantic_head(bev_feature_upscale)
        trajectory_query, agents_query = query_out.split(self._query_splits, dim=1)

        output: Dict[str, torch.Tensor] = {"bev_semantic_map": bev_semantic_map}
        trajectory = self._trajectory_head(trajectory_query)
        output.update(trajectory)

        agents = self._agent_head(agents_query)
        output.update(agents)

        return output


class AgentHead(nn.Module):
    def __init__(
        self,
        num_agents: int,
        d_ffn: int,
        d_model: int,
    ):
        super(AgentHead, self).__init__()

        self._num_objects = num_agents
        self._d_model = d_model
        self._d_ffn = d_ffn

        self._mlp_states = nn.Sequential(
            nn.Linear(self._d_model, self._d_ffn),
            nn.ReLU(),
            nn.Linear(self._d_ffn, BoundingBox2DIndex.size()),
        )

        self._mlp_label = nn.Sequential(
            nn.Linear(self._d_model, 1),
        )

    def forward(self, agent_queries) -> Dict[str, torch.Tensor]:

        agent_states = self._mlp_states(agent_queries)
        agent_states[..., BoundingBox2DIndex.POINT] = (
            agent_states[..., BoundingBox2DIndex.POINT].tanh() * 32
        )
        agent_states[..., BoundingBox2DIndex.HEADING] = (
            agent_states[..., BoundingBox2DIndex.HEADING].tanh() * np.pi
        )

        agent_labels = self._mlp_label(agent_queries).squeeze(dim=-1)

        return {"agent_states": agent_states, "agent_labels": agent_labels}


class TrajectoryHead(nn.Module):
    def __init__(self, num_poses: int, d_ffn: int, d_model: int):
        super(TrajectoryHead, self).__init__()

        self._num_poses = num_poses
        self._d_model = d_model
        self._d_ffn = d_ffn

        self._mlp = nn.Sequential(
            nn.Linear(self._d_model, self._d_ffn),
            nn.ReLU(),
            nn.Linear(self._d_ffn, num_poses * StateSE2Index.size()),
        )

    def forward(self, object_queries) -> Dict[str, torch.Tensor]:
        poses = self._mlp(object_queries).reshape(-1, self._num_poses, StateSE2Index.size())
        poses[..., StateSE2Index.HEADING] = poses[..., StateSE2Index.HEADING].tanh() * np.pi
        return {"trajectory": poses}