Draws a scene from multiple different modes. Args: im (array): the image to draw onto K (array): the 3x3 matrix for projecti

Args: x: Tensor of shape (N,C,H,W). H, W must be a multiple of ``self.size_divisibility``. Returns: dict[str

Input: cam_voxel_feat shape: (B, cam_C, voxel_z, voxel_y, voxel_x) point_feat shape: (B, point_C, voxel_z, voxel_x)

Forward function for `MultiheadAttention`. **kwargs allow passing a more general data flow when combining with other operations in `tr

Input: - srcs: List([bs, c, h, w]) - masks: List([bs, h, w])

Input: x shape: (B, in_c, H, W) Ks shape: (B, 3, 3)

Implements the multihead softmax attention. Arguments --------- q: The tensor containing the query. (B, T, H, D)

x: (batch, seqlen, hidden_dim) (where hidden_dim = num heads * head dim) key_padding_mask: bool tensor of shape (batch, seqlen)

Input: cam_bev_feat shape: (B, C, bev_y, bev_x) point_bev_feat: (B, C, bev_y, bev_x) update_cam_bev_feat:

Forward function.

:param query (N, Length_{query}, C) :param reference_points (N, Length_{query}, n_levels, 2), range

Forward function for `SinePositionalEncoding`. Args: mask (Tensor): ByteTensor mask. Non-zero values representing

Forward function for `LearnedPositionalEncoding`. Args: mask (Tensor): ByteTensor mask. Non-zero values representing

Forward function for `voxel pooling. Args: geom_xyz (Tensor): xyz coord for each voxel with the shape of [num_pro

Generalized IoU from https://giou.stanford.edu/ The boxes should be in [x0, y0, x1, y1] format Returns a [N, M] pairwise matrix, where

Description: One-to-one matching for a single image.