def __init__(self, checkpoint): #experiment_name='None', driver_conf=None, memory_fraction=0.18, #image_cut=[115, 510]): # use_planner=False,graph_file=None,map_file=None,augment_left_right=False,image_cut = [170,518]): Agent.__init__(self) # This should likely come from global #config_gpu = tf.ConfigProto() #config_gpu.gpu_options.visible_device_list = '0' #config_gpu.gpu_options.per_process_gpu_memory_fraction = memory_fraction #self._sess = tf.Session(config=config_gpu) # THIS DOES NOT WORK FOR FUSED PLUS LSTM #if self._config.number_frames_sequenced > self._config.number_frames_fused: # self._config_train.batch_size = self._config.number_frames_sequenced #else: # self._config_train.batch_size = self._config.number_frames_fused #self._train_manager = load_system(self._config_train) #self._config.train_segmentation = False self.model = CoILModel(g_conf.MODEL_NAME) self.model.load_state_dict(checkpoint['state_dict']) self.model.cuda() self.model.eval()
def __init__(self, city_name):
self.timer = Timer()
# Map setup
self._map = CarlaMap(city_name)
self._centerlines = Centerlines(city_name)
# Agent Setup
Agent.__init__(self)
self._neural_net = CAL_network()
self._seq_len = 5 #self._neural_net.model.max_input_shape
self._state = VehicleState()
self._agents_present = False
# Controller setup
param_path = os.path.dirname(__file__) + '/controller/params/'
cruise_params = get_params_from_txt(param_path + 'cruise_params.txt')
self._PID_cruise = PID(*cruise_params)
follow_params = get_params_from_txt(param_path + 'follow_params.txt')
self._PID_follow = PID(*follow_params)
# General Parameter setup
general_params = get_params_from_txt(param_path + 'general_params.txt')
self.c, self.d = general_params[0], general_params[1]
self.Kl_STANLEY = general_params[2]
self.Kr_STANLEY = general_params[3]
self.K0_STANLEY = general_params[4]
self.curve_slowdown = general_params[5]
self.DELTAl = general_params[6]
self.DELTAr = general_params[7]
self.DELTA0 = general_params[8]
self.EXP_DECAY = general_params[9]
def __init__(self,
city_name,
avoid_stopping,
memory_fraction=0.25,
image_cut=[115, 510]):
Agent.__init__(self)
scopeName = 'NET'
scopeName1 = 'First'
scopeName2 = 'Second'
self.dropout_vec = [1.0] * 8 + [0.7] * 2 + [0.5] * 2 + [0.5] * 1 + [
0.5, 1.
] * 2
config_gpu = tf.ConfigProto()
# GPU to be selected, just take zero , select GPU with CUDA_VISIBLE_DEVICES
config_gpu.gpu_options.visible_device_list = '0'
config_gpu.gpu_options.per_process_gpu_memory_fraction = memory_fraction
self._image_size = (88, 200, 3)
self._avoid_stopping = avoid_stopping
self._sess = tf.Session(config=config_gpu)
with tf.device('/gpu:0'):
self._input_images = tf.placeholder("float",
shape=[
None, self._image_size[0],
self._image_size[1],
self._image_size[2]
],
name="input_image")
self._input_data = tf.placeholder(tf.float32,
shape=[None, 1],
name="input_speed")
self._dout = tf.placeholder("float", shape=[len(self.dropout_vec)])
with tf.variable_scope(scopeName) as scope:
self._network_tensor = load_imitation_learning_network(
self._input_images, self._input_data, self._image_size,
self._dout, scopeName1, scopeName2)
import os
dir_path = os.path.dirname(__file__)
self._models_path = dir_path + '/model/test/'
# tf.reset_default_graph()
self._sess.run(tf.global_variables_initializer())
self.load_model()
self._image_cut = image_cut
def __init__(self,
city_name,
avoid_stopping,
memory_fraction=0.25,
image_cut=[115, 510]):
Agent.__init__(self)
self.dropout_vec = [1.0] * 8 + [0.7] * 2 + [0.5] * 2 + [0.5] * 1 + [
0.5, 1.
] * 5
config_gpu = tf.ConfigProto()
# GPU to be selected, just take zero , select GPU with CUDA_VISIBLE_DEVICES
# config_gpu.gpu_options.visible_device_list = '0'
# config_gpu.gpu_options.per_process_gpu_memory_fraction = memory_fraction
self._image_size = (88, 200, 3)
self._avoid_stopping = avoid_stopping
self._sess = tf.Session(config=config_gpu)
with tf.device('/cpu:0'):
self._input_images = tf.placeholder("float",
shape=[
None, self._image_size[0],
self._image_size[1],
self._image_size[2]
],
name="input_image")
self._input_data = []
self._input_data.append(
tf.placeholder(tf.float32,
shape=[None, 3],
name="target_control"))
self._input_data.append(
tf.placeholder(tf.float32, shape=[None, 1],
name="input_speed"))
self._dout = tf.placeholder("float", shape=[len(self.dropout_vec)])
# with tf.name_scope("Network"):
# self._network_tensor = make_network()
self._network_tensor = make_network()
dir_path = os.path.dirname(__file__)
self._models_path = dir_path + '/data_fine_tune/'
# tf.reset_default_graph()
self._sess.run(tf.global_variables_initializer())
self.load_model()
self._image_cut = image_cut
def __init__(self, checkpoint):
Agent.__init__(self)
self.checkpoint = checkpoint # We save the checkpoint for some interesting future use.
self.model = CoILModel(g_conf.MODEL_NAME)
self.model.load_state_dict(checkpoint['state_dict'])
self.model.cuda()
def __init__(self, checkpoint, architecture_name):
#experiment_name='None', driver_conf=None, memory_fraction=0.18,
#image_cut=[115, 510]):
# use_planner=False,graph_file=None,map_file=None,augment_left_right=False,image_cut = [170,518]):
Agent.__init__(self)
# This should likely come from global
#config_gpu = tf.ConfigProto()
#config_gpu.gpu_options.visible_device_list = '0'
#config_gpu.gpu_options.per_process_gpu_memory_fraction = memory_fraction
#self._sess = tf.Session(config=config_gpu)
# THIS DOES NOT WORK FOR FUSED PLUS LSTM
#if self._config.number_frames_sequenced > self._config.number_frames_fused:
# self._config_train.batch_size = self._config.number_frames_sequenced
#else:
# self._config_train.batch_size = self._config.number_frames_fused
#self._train_manager = load_system(self._config_train)
#self._config.train_segmentation = False
self.architecture_name = architecture_name
if architecture_name == 'coil_unit':
self.model_task, self.model_gen = CoILModel('coil_unit')
self.model_task, self.model_gen = self.model_task.cuda(
), self.model_gen.cuda()
elif architecture_name == 'unit_task_only':
self.model_task, self.model_gen = CoILModel('unit_task_only')
self.model_task, self.model_gen = self.model_task.cuda(
), self.model_gen.cuda()
else:
self.model = CoILModel(architecture_name)
self.model.cuda()
if architecture_name == 'wgangp_lsd':
# print(ckpt, checkpoint['best_loss_iter_F'])
self.model.load_state_dict(checkpoint['stateF_dict'])
self.model.eval()
elif architecture_name == 'coil_unit':
self.model_task.load_state_dict(checkpoint['task'])
self.model_gen.load_state_dict(checkpoint['b'])
self.model_task.eval()
self.model_gen.eval()
elif architecture_name == 'coil_icra':
self.model.load_state_dict(checkpoint['state_dict'])
self.model.eval()
elif architecture_name == 'unit_task_only':
self.model_task.load_state_dict(checkpoint['task_state_dict'])
self.model_gen.load_state_dict(checkpoint['enc_state_dict'])
self.model_task.eval()
self.model_gen.eval()
def __init__(self, city_name, avoid_stopping, memory_fraction=0.25, image_cut=[115, 510]):
Agent.__init__(self)
self._image_size = (88, 200, 3)
self._avoid_stopping = avoid_stopping
# Reading models and weights
dir_path = 'E:\GP\org data less'
self._model_path = dir_path + '/BH1_Nvidia.h5'
self._weights_path = dir_path + '/BH1_Nividia_at_epoch_40.h5'
self._image_cut = image_cut
self.model = keras.models.load_model(self._model_path, custom_objects={'masked_loss_function': masked_loss_function})
self.model.load_weights(self._weights_path)
def __init__(self,
city_name,
args_file='',
model_file='',
n_actions=0,
frameskip=1):
Agent.__init__(self)
self.args = self.read_args(args_file)
self.args.model = model_file
self.n_actions = n_actions
self.n_meas = self.compute_n_meas(self.args)
self.args.town_traintest = city_name
self.setup_model(self.n_actions, self.n_meas, self.args)
self.setup_data_preprocessor(self.args)
self.frameskip = frameskip
self.step = 0
def __init__(self,
city_name,
mode,
num_control,
path=None,
image_cut=[115, 510],
gpu_fraction=0.75):
Agent.__init__(self)
# set keras session
config_gpu = tf.ConfigProto()
config_gpu.gpu_options.allow_growth = True
config_gpu.gpu_options.per_process_gpu_memory_fraction = gpu_fraction
KTF.set_session(tf.Session(config=config_gpu))
self.model = None
self.mode = mode
self.num_control = num_control
self.path = path
self._image_cut = image_cut
self.init()
def __init__(self,
city_name,
avoid_stopping,
memory_fraction=0.25,
image_cut=[115, 510]):
Agent.__init__(self)
self.images = list([])
self.imagenum = 0
self._image_size = (88, 200, 3)
self._avoid_stopping = avoid_stopping
self.network = make_network()
self._sess = tf.Session(config=tf.ConfigProto(
log_device_placement=False))
self._sess.run(tf.global_variables_initializer())
saver = tf.train.Saver(write_version=saver_pb2.SaverDef.V2)
saver.restore(self._sess, './agents/imitation/mymodel/epoch-149.ckpt')
print('hellohellohellohellohellohello')
self._image_cut = image_cut
def __init__(self,
city_name,
task,
intersection,
save_choice,
avoid_stopping,
min_frames,
max_frames,
trajectory_no,
memory_fraction=0.25,
image_cut=[115, 510],
opt=0):
Agent.__init__(self)
self.opt = opt
self.save_choice = save_choice
self.intersection = intersection
task = task
self.trajectory_no = trajectory_no
self.dataset_pt = "datasets_" + str(
self.trajectory_no) + "/" + task + "/" + self.intersection + "/"
self.clean_frame_pt = self.dataset_pt + "clean_frame/"
self.input_pt = self.dataset_pt + "inputInfo/"
self.images_save_path = "out_" + self.intersection + "_" + str(
self.trajectory_no) + "/"
self.steer_pt = self.dataset_pt + "steer/"
self.steer_deviation_clean_pt = self.dataset_pt + "steer_deviation_cleanFrame/"
self.steer_pt_input = self.dataset_pt + "steer_input/"
self.control_info_pt = self.dataset_pt + "control_input/"
self.frame_count = 0
self.min_frames = min_frames
self.max_frames = max_frames
if not os.path.exists(self.dataset_pt):
os.mkdir(self.dataset_pt)
if not os.path.exists(self.input_pt):
os.mkdir(self.input_pt)
if not os.path.exists(self.steer_pt):
os.mkdir(self.steer_pt)
if not os.path.exists(self.steer_deviation_clean_pt):
os.mkdir(self.steer_deviation_clean_pt)
self.dropout_vec = [1.0] * 8 + [0.7] * 2 + [0.5] * 2 + [0.5] * 1 + [
0.5, 1.
] * 5
config_gpu = tf.ConfigProto()
# GPU to be selected, just take zero , select GPU with CUDA_VISIBLE_DEVICES
config_gpu.gpu_options.visible_device_list = '0'
config_gpu.gpu_options.per_process_gpu_memory_fraction = memory_fraction
self._image_size = (88, 200, 3)
self._avoid_stopping = avoid_stopping
""" loading adversary """
self.city_name = city_name
self._sess = tf.Session(config=config_gpu)
with tf.device('/cpu:0'):
self._input_images = tf.placeholder("float",
shape=[
None, self._image_size[0],
self._image_size[1],
self._image_size[2]
],
name="input_image")
self._input_data = []
self.start = True
self.frame_count = 0
self._input_data.append(
tf.placeholder(tf.float32,
shape=[None, 4],
name="input_control"))
self._input_data.append(
tf.placeholder(tf.float32, shape=[None, 1],
name="input_speed"))
self._dout = tf.placeholder("float", shape=[len(self.dropout_vec)])
with tf.name_scope("Network"):
self._network_tensor, self.gradients_steers, self.gradients_acc, self.gradients_brake = load_imitation_learning_network(
self._input_images, self._input_data, self._image_size,
self._dout)
dir_path = os.path.dirname(__file__)
self._models_path = dir_path + '/model/'
# tf.reset_default_graph()
self._sess.run(tf.global_variables_initializer())
self.load_model()
self._image_cut = image_cut
def __init__(self,
city_name,
avoid_stopping,
gpu_num=0,
memory_fraction=0.25,
image_cut=[115, 510]):
"""
Source: https://github.com/carla-simulator/imitation-learning
"""
Agent.__init__(self)
self._image_size = (88, 200, 3)
self._avoid_stopping = avoid_stopping
self.dropout_vec = [1.0] * 8 + [0.7] * 2 + [0.5] * 2 + [0.5] * 1 + [
0.5, 1.
] * 5
config_gpu = tf.ConfigProto()
#print(tf.test.is_gpu_available())
#g = input()
#if tf.test.is_gpu_available():
# tf_device = '/gpu:' + str(gpu_num)
# # GPU to be selected, just take zero , select GPU with CUDA_VISIBLE_DEVICES
# config_gpu.gpu_options.visible_device_list = '0'
# config_gpu.gpu_options.visible_device_list = str(gpu_num)
#else:
# tf_device = '/cpu:0'
tf_device = '/cpu:0'
config_gpu.gpu_options.visible_device_list = '0'
config_gpu.gpu_options.per_process_gpu_memory_fraction = memory_fraction
self._sess = tf.Session(config=config_gpu)
with tf.device(tf_device):
self._input_images = tf.placeholder("float",
shape=[
None, self._image_size[0],
self._image_size[1],
self._image_size[2]
],
name="input_image")
self._input_data = []
self._input_data.append(
tf.placeholder(tf.float32,
shape=[None, 4],
name="input_control"))
self._input_data.append(
tf.placeholder(tf.float32, shape=[None, 1],
name="input_speed"))
self._dout = tf.placeholder("float", shape=[len(self.dropout_vec)])
with tf.name_scope("Network"):
self._network_tensor = load_imitation_learning_network(
self._input_images, self._input_data, self._image_size,
self._dout)
import os
dir_path = os.path.dirname(__file__)
self._models_path = dir_path + '/model/'
# tf.reset_default_graph()
self._sess.run(tf.global_variables_initializer())
self.load_model()
self._image_cut = image_cut
