def infer_map(
test_loader: torch.utils.data.DataLoader,
encoder: EncoderRNN,
decoder: DecoderRNN,
start_idx: int,
forecasted_save_dir: str,
model_utils: ModelUtils,
):
"""Infer function for map-based LSTM baselines and save the forecasted trajectories.
Args:
test_loader: DataLoader for the test set
encoder: Encoder network instance
decoder: Decoder network instance
start_idx: start index for the current joblib batch
forecasted_save_dir: Directory where forecasted trajectories are to be saved
model_utils: ModelUtils instance
"""
args = parse_arguments()
global best_loss
forecasted_trajectories = {}
for i, (_input, target, helpers) in enumerate(test_loader):
_input = _input.to(device)
batch_helpers = list(zip(*helpers))
helpers_dict = {}
for k, v in config.LSTM_HELPER_DICT_IDX.items():
helpers_dict[k] = batch_helpers[v]
# Set to eval mode
encoder.eval()
decoder.eval()
# Encoder
batch_size = _input.shape[0]
input_length = _input.shape[1]
# Iterate over every element in the batch
for batch_idx in range(batch_size):
num_candidates = len(
helpers_dict["CANDIDATE_CENTERLINES"][batch_idx])
curr_centroids = helpers_dict["CENTROIDS"][batch_idx]
seq_id = int(helpers_dict["SEQ_PATHS"][batch_idx])
abs_outputs = []
# Predict using every centerline candidate for the current trajectory
for candidate_idx in range(num_candidates):
curr_centerline = helpers_dict["CANDIDATE_CENTERLINES"][
batch_idx][candidate_idx]
curr_nt_dist = helpers_dict["CANDIDATE_NT_DISTANCES"][
batch_idx][candidate_idx]
_input = torch.FloatTensor(
np.expand_dims(curr_nt_dist[:args.obs_len].astype(float),
0)).to(device)
# Initialize encoder hidden state
encoder_hidden = model_utils.init_hidden(
1, encoder.module.hidden_size
if use_cuda else encoder.hidden_size)
# Encode observed trajectory
for ei in range(input_length):
encoder_input = _input[:, ei, :]
encoder_hidden = encoder(encoder_input, encoder_hidden)
# Initialize decoder input with last coordinate in encoder
decoder_input = encoder_input[:, :2]
# Initialize decoder hidden state as encoder hidden state
decoder_hidden = encoder_hidden
decoder_outputs = torch.zeros((1, args.pred_len, 2)).to(device)
# Decode hidden state in future trajectory
for di in range(args.pred_len):
decoder_output, decoder_hidden = decoder(
decoder_input, decoder_hidden)
decoder_outputs[:, di, :] = decoder_output
# Use own predictions as inputs at next step
decoder_input = decoder_output
# Get absolute trajectory
abs_helpers = {}
abs_helpers["REFERENCE"] = np.expand_dims(
np.array(helpers_dict["CANDIDATE_DELTA_REFERENCES"]
[batch_idx][candidate_idx]),
0,
)
abs_helpers["CENTERLINE"] = np.expand_dims(curr_centerline, 0)
abs_input, abs_output = baseline_utils.get_abs_traj(
_input.clone().cpu().numpy(),
decoder_outputs.detach().clone().cpu().numpy(),
args,
abs_helpers,
)
# array of shape (1,30,2) to list of (30,2)
abs_outputs.append(abs_output[0])
forecasted_trajectories[seq_id] = abs_outputs
os.makedirs(forecasted_save_dir, exist_ok=True)
with open(os.path.join(forecasted_save_dir, f"{start_idx}.pkl"),
"wb") as f:
pkl.dump(forecasted_trajectories, f)
def infer_and_save_traj_absolute(
self,
grid_search: Any,
train_output: np.ndarray,
test_input: np.ndarray,
test_references: Union[np.ndarray, None],
test_seq_ids: np.ndarray,
test_translation: Union[np.ndarray, None],
test_rotation: Union[np.ndarray, None],
start_idx: int,
args: Any,
num_features: int,
horizon: int,
save_dir: str,
) -> None:
"""Non-map baselines inference. This function does the inference based on the given model, and saves the forecasted trajectories.
Args:
grid_search: GridSearchCV object
train_output: Train output
test_input: Test input
test_references: References for reverting delta transformation of candidate trajectories,
test_seq_ids: csv name for the sequence,
test_translation: Translation used for normalizing trajectories,
test_rotation: Rotation used for normalizing trajectories,
start_idx: start_idx for current joblib batch,
args: Arguments passed to the baseline script,
num_features: Number of features,
horizon: Prediction Horizon,
save_dir: Directory where forecasted trajectories are to be saved
"""
test_num_tracks = test_input.shape[0]
print(f"Absolute Inference currently at index {start_idx} ...")
forecasted_trajectories = {}
# Preprocess and get neighbors
pipeline_steps = grid_search.best_estimator_.named_steps.keys()
preprocessed_test_input = np.copy(test_input)
for step in pipeline_steps:
curr_step = grid_search.best_estimator_.named_steps[step]
# Get neighbors
if step == "regressor":
neigh_idx = curr_step.kneighbors(
preprocessed_test_input,
return_distance=False,
n_neighbors=args.n_neigh,
)
# Preprocess
else:
if curr_step is not None:
preprocessed_test_input = curr_step.transform(
preprocessed_test_input)
# Predict for each trajectory
for i in range(test_num_tracks):
curr_test_input = np.repeat(test_input[i],
repeats=args.n_neigh,
axis=0)
neigh_output = train_output[neigh_idx][
i, :, :horizon, :] # num_neigh x curr_pred_len x 2
abs_helpers = {}
if args.use_delta:
abs_helpers["REFERENCE"] = np.array(
[test_references[i] for _ in range(args.n_neigh)])
if args.normalize:
abs_helpers["TRANSLATION"] = np.array(
[test_translation[i] for _ in range(args.n_neigh)])
abs_helpers["ROTATION"] = np.array(
[test_rotation[i] for _ in range(args.n_neigh)])
# Convert trajectory to map frame
abs_input, abs_output = baseline_utils.get_abs_traj(
curr_test_input.copy().reshape(
(-1, args.obs_len, num_features), order="F"),
neigh_output.copy(),
args,
helpers=abs_helpers,
)
forecasted_trajectories[test_seq_ids[i]] = abs_output
with open(f"{save_dir}/{start_idx}.pkl", "wb") as f:
pkl.dump(forecasted_trajectories, f)
def infer_absolute(
test_loader: torch.utils.data.DataLoader,
encoder: EncoderRNN,
decoder: DecoderRNN,
start_idx: int,
forecasted_save_dir: str,
model_utils: ModelUtils,
):
"""Infer function for non-map LSTM baselines and save the forecasted trajectories.
Args:
test_loader: DataLoader for the test set
encoder: Encoder network instance
decoder: Decoder network instance
start_idx: start index for the current joblib batch
forecasted_save_dir: Directory where forecasted trajectories are to be saved
model_utils: ModelUtils instance
"""
args = parse_arguments()
forecasted_trajectories = {}
for i, (_input, target, helpers) in enumerate(test_loader):
_input = _input.to(device)
batch_helpers = list(zip(*helpers))
helpers_dict = {}
for k, v in config.LSTM_HELPER_DICT_IDX.items():
helpers_dict[k] = batch_helpers[v]
# Set to eval mode
encoder.eval()
decoder.eval()
# Encoder
batch_size = _input.shape[0]
input_length = _input.shape[1]
input_shape = _input.shape[2]
# Initialize encoder hidden state
encoder_hidden = model_utils.init_hidden(
batch_size,
encoder.module.hidden_size if use_cuda else encoder.hidden_size)
# Encode observed trajectory
for ei in range(input_length):
encoder_input = _input[:, ei, :]
encoder_hidden = encoder(encoder_input, encoder_hidden)
# Initialize decoder input with last coordinate in encoder
decoder_input = encoder_input[:, :2]
# Initialize decoder hidden state as encoder hidden state
decoder_hidden = encoder_hidden
decoder_outputs = torch.zeros(
(batch_size, args.pred_len, 2)).to(device)
# Decode hidden state in future trajectory
for di in range(args.pred_len):
decoder_output, decoder_hidden = decoder(decoder_input,
decoder_hidden)
decoder_outputs[:, di, :] = decoder_output
# Use own predictions as inputs at next step
decoder_input = decoder_output
# Get absolute trajectory
abs_helpers = {}
abs_helpers["REFERENCE"] = np.array(helpers_dict["DELTA_REFERENCE"])
abs_helpers["TRANSLATION"] = np.array(helpers_dict["TRANSLATION"])
abs_helpers["ROTATION"] = np.array(helpers_dict["ROTATION"])
abs_inputs, abs_outputs = baseline_utils.get_abs_traj(
_input.clone().cpu().numpy(),
decoder_outputs.detach().clone().cpu().numpy(),
args,
abs_helpers,
)
for i in range(abs_outputs.shape[0]):
seq_id = int(helpers_dict["SEQ_PATHS"][i])
forecasted_trajectories[seq_id] = [abs_outputs[i]]
with open(os.path.join(forecasted_save_dir, f"{start_idx}.pkl"),
"wb") as f:
pkl.dump(forecasted_trajectories, f)
def infer_and_save_traj_map(
self,
grid_search: Any,
train_output: np.ndarray,
test_nt: np.ndarray,
test_centerlines: np.ndarray,
test_references: np.ndarray,
test_seq_ids: np.ndarray,
start_idx: int,
args: Any,
num_features: int,
horizon: int,
save_dir: str,
) -> np.ndarray:
"""Map-based baselines inference. This function does the inference based on the given model, and saves the forecasted trajectories.
Args:
grid_search: GridSearchCV object,
train_output: Train output of shape [num_tracks, pred_len, num_features]
test_nt: Candidate trajectories in centerline curvilinear coordinates,
test_centerlines: Candidate centerlines,
test_references: References for reverting delta transformation of candidate trajectories,
test_seq_ids: csv name for the sequence,
start_idx: start_idx for current joblib batch,
args: Arguments passed to the baseline script,
num_features: Number of features,
horizon: Prediction Horizon,
save_dir: Directory where forecasted trajectories are to be saved
Returns:
forecasted_trajectories: Forecasted trajectories
"""
test_num_tracks = test_nt.shape[0]
print(f"Map-based Inference currently at index {start_idx} ...")
forecasted_trajectories = {}
# Predict for each trajectory
for i in range(test_num_tracks):
# Helpers for current track
test_nt_i = test_nt[i]
test_cl_i = test_centerlines[i]
test_references_i = test_references[i]
curr_forecasted_trajectories = []
# Predict using each of candidate centerlines
for (
test_nt_i_curr_candidate,
test_cl_i_curr_candidate,
test_ref_i_curr_candidate,
) in zip(test_nt_i, test_cl_i, test_references_i):
test_input = test_nt_i_curr_candidate[:args.
obs_len, :].reshape(
(1,
2 * args.obs_len),
order="F")
# Preprocess and get neighbors
pipeline_steps = grid_search.best_estimator_.named_steps.keys()
preprocessed_test_input = np.copy(test_input)
for step in pipeline_steps:
curr_step = grid_search.best_estimator_.named_steps[step]
# Get neighbors
if step == "regressor":
neigh_idx = curr_step.kneighbors(
preprocessed_test_input,
return_distance=False,
n_neighbors=args.n_neigh,
)
# Preprocess
else:
if curr_step is not None:
print(preprocessed_test_input.shape)
preprocessed_test_input = curr_step.transform(
preprocessed_test_input)
y_pred = train_output[neigh_idx][
0, :, :horizon, :] # num_neighbors x curr_pred_len x 2
test_input = np.repeat(test_input,
repeats=args.n_neigh,
axis=0)
abs_helpers = {}
abs_helpers["CENTERLINE"] = [
test_cl_i_curr_candidate for i in range(args.n_neigh)
]
if args.use_delta:
abs_helpers["REFERENCE"] = np.array([
test_ref_i_curr_candidate for i in range(args.n_neigh)
])
# Convert trajectory to map frame
abs_input, abs_output = baseline_utils.get_abs_traj(
test_input.copy().reshape((-1, args.obs_len, num_features),
order="F"),
y_pred.copy(),
args,
helpers=abs_helpers,
)
curr_forecasted_trajectories.extend(abs_output)
forecasted_trajectories[
test_seq_ids[i]] = curr_forecasted_trajectories
with open(f"{save_dir}/{start_idx}.pkl", "wb") as f:
pkl.dump(forecasted_trajectories, f)
