- x : (B, T, D) - map_grid_feat : (B, C, H, W) - raster_from_agent: (B, 3, 3)

Reshape selected dimensions for all tensors in nested dictionary or list or tuple to a target dimension. Args: x (dict or list o

added for scenediffuser - net_in : (B, M, T, 8) - net_out : (B*M, d_k)

Verify and expand map to the number of necessary channels if necessary.

Log probability of tensor x under a uniform mixture of Gaussians. Adapted from CS 236 at Stanford. Args: x (torch.Tensor): tensor

Filter agents according to a set of rules: if new agent (not in tracked_agents) then: - must be a car - must be in dis

Get visualisation objects for the current frame. :param mapAPI: mapAPI object (used for lanes, crosswalks etc..) :param frame: the current fr

Rasterise agents of interest for a given frame in a given scene. :param scene_index: index of the scene :param state_index: frame ind

Applies hard constraints to positions (x,y) specified by the given configuration. - x : trajectory to update with constraints. (B, N, T, D) w

Factory function for vectorizers, reads the config, loads required data and initializes the vectorizer. Args: cfg (dict): Config.

Get a combined observation by concatenating ego and agents

Compute KL Divergence loss Args: outputs (dict): outputs of the self.forward() call Returns: a dict

Evaluate progress towards a goal that we want to hit. - pos_pred : (B x N x T x 2) - tgt_pos : (B x 2) - urgency : (B) in (0.0, 1.0]

- batch_maps (B, C, H, W)

This will convert 9 fields (some of which only have center values) from scene-centric to agent-centric and return a new dict consisting only of t

cosine schedule as proposed in https://openreview.net/forum?id=-NEXDKk8gZ

Cast coordinates to numpy.int but keep fractional part by previously multiplying by 2**CV2_SHIFT cv2 calls will use shift to restore original

Computes whether a collision occured between ego and any another agent. Also computes the type of collision: rear, front, or side. For th

whether the map needs to be rendered

A basic cost function for prediction-and-planning

Flatten all tensors in nested dictionary or list or tuple, from @begin_axis onwards. Args: x (dict or list or tuple): a possibly nes

This function indexes out a target dimension of a tensor in a structured way, by allowing a different value to be selected for each member of

For masking out the subsequent info.

Predict an action based on the input observation

PERCEPTION_LABELS = [ "PERCEPTION_LABEL_NOT_SET", "PERCEPTION_LABEL_UNKNOWN", "PERCEPTION_LABEL_DONTCARE", "PERCEPTION_LABEL_CAR"

Output: lanes_points: (S_seg, S_point, 4)

Get veh-veh and veh-ped collision loss.

Cost for road departure.