P2Trainer

module P2Trainer
class P2Trainer
The Precision-Level 2 (P2) Trainer class object instantiates a training
session using PyTorch's MaskRCNN-Benchmark for taking an input dataset and
fasterrcnn model from PyTorch model zoo in order to generate a
custom-trained P2 ONNX model file.\n
This model can then be deployed as a ROS2 package.

Public Functions

__init__(self, path_to_dataset, model_name, label_list, max_iteration, checkpoint_period, test_period, steps)
The constructor.
Sets all the required fixed path to various files needed to start
a training session.\n
Calls updateTrainingConfig function.
updateTrainingConfig(self)
A Mutator function that modifies the various training session config
files.
updateTrainVerification(self)
checkGPUAvailability(self)
train(self, debug)
export(self, debug)
isTrainFarmSetupVerified(self)
isExporterSetupVerified(self)
pullTrainFarmDockerImage(self)
createTrainFarmDockerContainer(self)
pullExporterDockerImage(self)
installTrainingDependencies(self)
createExportDockerContainer(self)
installExporterDependencies(self)
copyTrainingFiles(self)
runTraining(self)
copyExportFiles(self)
runExporter(self)

Public Members

p2_train_logger
model_name
label_list
max_iteration
checkpoint_period
test_period
steps
isGPUAvailableFlag
path_to_dataset
path_to_modif
path_to_training_config
path_to_export_config
path_to_trim_tools
path_to_remove_init_tool
path_to_export_modif
docker_inspect_process
docker_pull_process
docker_construct_process
install_depend_process
copy_process
training_process
exporter_process
shift_process

Private Members

_TRAIN_DOCKER_IMG
_TRAIN_DOCKER_CONTAINER
_EXPORT_DOCKER_IMG
_EXPORT_DOCKER_CONTAINER