lingvo.tasks.car.params.kitti module

Train models on KITTI data.

lingvo.tasks.car.params.kitti.KITTITrainSpec(params)

lingvo.tasks.car.params.kitti.KITTIValSpec(params)

lingvo.tasks.car.params.kitti.KITTITestSpec(params)

class lingvo.tasks.car.params.kitti.KITTITrain(*args, **kwargs)

Bases: lingvo.tasks.car.kitti_input_generator.KITTILaser

KITTI train set with raw laser data.

classmethod Params()

Defaults params.

class lingvo.tasks.car.params.kitti.KITTIValidation(*args, **kwargs)

Bases: lingvo.tasks.car.kitti_input_generator.KITTILaser

KITTI validation set with raw laser data.

classmethod Params()

Defaults params.

class lingvo.tasks.car.params.kitti.KITTITest(*args, **kwargs)

Bases: lingvo.tasks.car.kitti_input_generator.KITTILaser

KITTI test set with raw laser data.

classmethod Params()

Defaults params.

class lingvo.tasks.car.params.kitti.KITTIGridTrain(*args, **kwargs)

Bases: lingvo.tasks.car.kitti_input_generator.KITTIGrid

KITTI train set with grid laser data.

classmethod Params()

Defaults params.

class lingvo.tasks.car.params.kitti.KITTIGridValidation(*args, **kwargs)

Bases: lingvo.tasks.car.kitti_input_generator.KITTIGrid

KITTI validation set with grid laser data.

classmethod Params()

Defaults params.

class lingvo.tasks.car.params.kitti.KITTIGridTest(*args, **kwargs)

Bases: lingvo.tasks.car.kitti_input_generator.KITTIGrid

KITTI validation set with grid laser data.

classmethod Params()

Defaults params.

class lingvo.tasks.car.params.kitti.KITTISparseLaserTrain(*args, **kwargs)

Bases: lingvo.tasks.car.kitti_input_generator.KITTISparseLaser

KITTI train set with sparse laser data.

classmethod Params()

Defaults params.

class lingvo.tasks.car.params.kitti.KITTISparseLaserValidation(*args, **kwargs)

Bases: lingvo.tasks.car.kitti_input_generator.KITTISparseLaser

KITTI validation set with sparse laser data.

classmethod Params()

Defaults params.

class lingvo.tasks.car.params.kitti.KITTISparseLaserTest(*args, **kwargs)

Bases: lingvo.tasks.car.kitti_input_generator.KITTISparseLaser

KITTI test set with sparse laser data.

classmethod Params()

Defaults params.

lingvo.tasks.car.params.kitti._MaybeRemove(values, key)

Remove the entry ‘key’ from ‘values’ if present.

lingvo.tasks.car.params.kitti.AddLaserAndCamera(params)

Adds laser and camera extractors.

class lingvo.tasks.car.params.kitti.StarNetCarsBase

Bases: lingvo.core.base_model_params.SingleTaskModelParams

StarNet model for cars.

RUN_LOCALLY = False

NUM_ANCHOR_BBOX_OFFSETS = 25

NUM_ANCHOR_BBOX_ROTATIONS = 4

NUM_ANCHOR_BBOX_DIMENSIONS = 1

FOREGROUND_ASSIGNMENT_THRESHOLD = 0.6

BACKGROUND_ASSIGNMENT_THRESHOLD = 0.45

INCLUDED_CLASSES = ['Car']

class AnchorBoxSettings

Bases: lingvo.tasks.car.input_preprocessors.SparseCarV1AnchorBoxSettings

ROTATIONS = [0, 1.5707963267948966, 2.356194490192345, 0.7853981633974483]

_configure_input(p)

Base function managing the delegation of job specific input configs.

_configure_generic_input(p)

Update input_config p for all jobs.

_configure_trainer_input(p)

Update input_config p for jobs running training.

_configure_decoder_input(p)

Update input_config p for jobs running decoding.

_configure_evaler_input(p)

Update input_config p for jobs running evaluation.

Train()

Returns Params for the training dataset.

Test()

Returns Params for the testing dataset.

Dev()

Returns Params for the development dataset.

Task()

Returns task params.

class lingvo.tasks.car.params.kitti.StarNetCarModel0701

Bases: lingvo.tasks.car.params.kitti.StarNetCarsBase

StarNet Car model trained on KITTI.

class AnchorBoxSettings

Bases: lingvo.tasks.car.input_preprocessors.SparseCarV1AnchorBoxSettings

CENTER_X_OFFSETS = array([-1.294, -0.647, 0. , 0.647, 1.294])

CENTER_Y_OFFSETS = array([-1.294, -0.647, 0. , 0.647, 1.294])

_configure_generic_input(p)

Update input_config p for all jobs.

_configure_trainer_input(p)

Update input_config p for jobs running training.

Task()

Returns task params.

class lingvo.tasks.car.params.kitti.StarNetPedCycModel0704

Bases: lingvo.tasks.car.params.kitti.StarNetCarsBase

StarNet Ped/Cyc model trained on KITTI.

INCLUDED_CLASSES = ['Pedestrian', 'Cyclist']

FOREGROUND_ASSIGNMENT_THRESHOLD = 0.48

BACKGROUND_ASSIGNMENT_THRESHOLD = 0.8

NUM_ANCHOR_BBOX_OFFSETS = 9

NUM_ANCHOR_BBOX_ROTATIONS = 4

NUM_ANCHOR_BBOX_DIMENSIONS = 3

class AnchorBoxSettings

Bases: lingvo.tasks.car.input_preprocessors.SparseCarV1AnchorBoxSettings

DIMENSION_PRIORS = [(0.6, 0.8, 1.7), (0.6, 0.6, 1.2), (0.6, 1.76, 1.73)]

ROTATIONS = [0, 1.5707963267948966, 2.356194490192345, 0.7853981633974483]

CENTER_X_OFFSETS = array([-0.31, 0. , 0.31])

CENTER_Y_OFFSETS = array([-0.31, 0. , 0.31])

CENTER_Z_OFFSETS = [-0.6]

_configure_generic_input(p)

Update input_config p for all jobs.

_configure_trainer_input(p)

Update input_config p for jobs running training.

_configure_decoder_input(p)

Update input_config p for jobs running decoding.

_configure_evaler_input(p)

Update input_config p for jobs running evaluation.

Task()

Returns task params.