def main():
""" Reads command line arguments and starts either training, evaluation or inference. """
parser = argparse.ArgumentParser()
parser.add_argument(
'--config',
default='',
help='JSON-formatted file with configuration parameters')
parser.add_argument('--mode',
default='',
help='The operational mode - train|evaluate|predict')
parser.add_argument('--input',
default='',
help='The path to a PLY file to run through the model')
parser.add_argument('--device',
default='gpu',
help='The device to use - cpu|gpu')
parser.add_argument('--colors',
default='',
help='A file containing colors')
cla = parser.parse_args()
if cla.config == '':
print 'Error: --config flag is required. Enter the path to a configuration file.'
if cla.mode == 'evaluate':
inference.evaluate(cla.config, cla.device)
elif cla.mode == 'predict':
inference.predict(cla.config, cla.input, cla.device, cla.colors)
elif cla.mode == 'train':
training.train(cla.config)
else:
print 'Error: invalid operational mode - options: train, evaluate or predict'
def on_message(client, userdata, msg):
print("topic: " + msg.topic)
print("Payload Data: " + str(msg.payload))
if str(msg.payload) == 'bottle':
os.chdir("./../pointnet")
location, std = evaluate(label_to_detect=12,
x_offset=0.35,
y_offset=0.137,
z_offset=0.1)
print(location, std)
def train(target_dir,
embedding_dim,
hidden_dim,
glove_file):
torch.manual_seed(1)
train_word_to_ix, train_tag_to_ix, train_sents_idx, train_labels_idx = pickle.load(
open(target_dir + "CoNLL_train.pkl", "rb"))
test_word_to_ix, test_tag_to_ix, test_sents_idx, test_labels_idx = pickle.load(
open(target_dir + "CoNLL_test.pkl", "rb"))
model = LSTMTagger(embedding_dim,
hidden_dim,
len(train_word_to_ix),
len(train_tag_to_ix),
target_dir,
glove_file)
criterion = nn.NLLLoss()
optimizer = optim.RMSprop(model.parameters())
EPOCHS = 2
for epoch in range(EPOCHS):
loss = 0
for i, (sentence, tags) in tqdm(enumerate(zip(train_sents_idx, train_labels_idx))):
model.zero_grad()
model.hidden = model.init_hidden()
# 単語インデックスの tensor に変換
sentence_in = utils.prepare_sequence(sentence)
# Tags インデックスの tensor に変換
targets = utils.prepare_sequence(tags)
tag_scores = model.forward(sentence_in)
loss = criterion(tag_scores, targets)
loss.backward()
optimizer.step()
loss += loss.item()
f1_score_train_sents_avg = inference.evaluate(model,
train_sents_idx[:len(test_sents_idx)],
train_labels_idx[:len(test_sents_idx)])
f1_score_test_sents_avg = inference.evaluate(model,
test_sents_idx,
test_labels_idx)
print("[{}] EPOCH {} - LOSS: {:.8f} TRAIN_DATA_F1_SCORE: {} TEST_DATA_F1_SCORE: {}".
format(datetime.datetime.today(), epoch + 1, loss,
f1_score_train_sents_avg, f1_score_test_sents_avg))
def AutoImageCaption():
image_url = request.args.get('image')
print('image_url')
image_extension = image_url[-4:]
image_path = tf.keras.utils.get_file(str(int(time.time())) +
image_extension,
origin=image_url)
result, attention_plot = evaluate(image_path)
data = {'Prediction Caption:': ' '.join(result)}
return jsonify(data)
def main():
config = Config()
# Load images and ground truth planes
data = input_data.read_data_sets(config.data_dir, config.label_dir,
config.train_list_file,
config.test_list_file,
config.landmark_count, config.plane_name)
print("Start inference...")
tf.reset_default_graph()
sess = tf.InteractiveSession()
# Load trained model
g = tf.get_default_graph()
saver = tf.train.import_meta_graph(
tf.train.latest_checkpoint(config.model_dir) + '.meta')
saver.restore(sess, tf.train.latest_checkpoint(config.model_dir))
action_prob_tran = g.get_collection('action_prob_tran')[
0] # translation classification probability
ytr = g.get_collection('ytr')[
0] # translation regression (displacement vector)
action_prob_rot = g.get_collection('action_prob_rot')[
0] # rotation classification probability
yrr_norm = g.get_collection('yrr_norm')[
0] # rotation regression (quaternions)
x = g.get_collection('x')[0]
keep_prob = g.get_collection('keep_prob')[0]
# Evaluation on test-set
print("Evaluation on test set:")
inference.evaluate(data.test, config, 'test', sess, x, action_prob_tran,
ytr, action_prob_rot, yrr_norm, keep_prob)
# Evaluation on train-set
print("Evaluation on train set:")
inference.evaluate(data.train, config, 'train', sess, x, action_prob_tran,
ytr, action_prob_rot, yrr_norm, keep_prob)
sess.close()
def main():
try:
print "============ Press `Enter` to recognize the object ..."
raw_input()
location, std = evaluate(label_to_detect=12,
x_offset=0,
y_offset=0,
z_offset=1.55)
print(location)
print(std)
#location=[1.0653, -0.3668, 1.7145]
print "============ Press `Enter` to set up the moveit_commander ..."
raw_input()
tutorial = MoveGroupPythonIntefaceTutorial()
print "============ Press `Enter` to add a box ..."
raw_input()
tutorial.add_box(location)
print "============ Press `Enter` to execute a movement using a pose goal ..."
raw_input()
tutorial.go_to_pose_goal(location)
print "============ Press `Enter` to attach the object to the robot ..."
raw_input()
tutorial.attach_box()
print "============ Press `Enter` to move back to initial position ..."
raw_input()
tutorial.go_to_joint_state()
print "============ Press `Enter` to dettach the object ..."
raw_input()
tutorial.detach_box()
print "============ Complete!"
raw_input()
tutorial.remove_box()
except rospy.ROSInterruptException:
return
except KeyboardInterrupt:
return
def evaluate(self, *args, **kwargs):
''' Evaluate the performance of the model. See inference.evaluate
for more details. '''
from inference import evaluate
return evaluate(self, *args, **kwargs)
from inference import evaluate
#label and their number
# 0 ceiling
# 1 floor
# 2 wall
# 3 beam
# 4 column
# 5 window
# 6 door
# 7 table
# 8 chair
# 9 sofa
# 10 bookcase
# 11 board
# 12 clutter
#----------------------------------------------------------------------------
# this is usage example of pointnet recognition in gazebo with kinect
#----------------------------------------------------------------------------
location, std = evaluate(label_to_detect=12,
x_offset=0.35,
y_offset=0.137,
z_offset=0.1)
#you will have mean location std of the object belong to clutter class
#if you run this on tensorflow-cpu, it will be slow when doing
#neural network inference, proberly 60 secs.
print(location)
print(std)
import subprocess
import sys
#label and their number
# 0 ceiling
# 1 floor
# 2 wall
# 3 beam
# 4 column
# 5 window
# 6 door
# 7 table
# 8 chair
# 9 sofa
# 10 bookcase
# 11 board
# 12 clutter
#----------------------------------------------------------------------------
# this is usage example of pointnet recognition in gazebo with kinect
#----------------------------------------------------------------------------
location, std = evaluate(label_to_detect=12,
x_offset=0,
y_offset=0,
z_offset=0)
#you will have mean location std of the object belong to clutter class
#if you run this on tensorflow-cpu, it will be slow when doing
#neural network inference, proberly 60 secs.
print(location)
print(std)