def run(bump):
while True:
command = raw_input("Command: ")
if command == 'q':
exit(0)
elif command == 'c':
print "right:", bump.getRightArmCoords()
print "left:", bump.getLeftArmCoords()
elif command[0] == 'l':
bump.moveLeftArmTo(make_tuple(command[1:]))
elif command[0] == 'r':
bump.moveRightArmTo(make_tuple(command[1:]))
elif command[0] == 'b':
coords = make_tuple(command[2:])
if command[1] =='r':
bump.bumpRight(coords)
elif command[1] =='l':
bump.bumpLeft(coords)
else:
print "Unknown command"
elif command[0] == 'w':
angle = make_tuple(command[2:])
if command[1] == 'l':
bump.rotateLeftWristTo(angle)
elif command[1] == 'r':
bump.rotateRightWristTo(angle)
else:
print "Unknown command"
else:
print "Unknown command"
def play(self):
print("How to play:\nExample: (x,y,z) to (x1,y1,z1)\nExample: KP0 to QP1")
while True:
self.board.pretty_print()
do_move = True
while do_move:
(a, b) = input("White's Move:").split(" to ")
try:
if a[0] in ("K", "Q"):
self.board.move_atk_board(a, b)
else:
self.board.move(make_tuple(a), make_tuple(b))
do_move = False
except InvalidMoveException as err:
print(err)
do_move = True
self.board.pretty_print()
while do_move:
(a, b) = input("Black's Move:").split(" to ")
try:
if a[0] in ("K", "Q"):
self.board.move_atk_board(a, b)
else:
self.board.move(make_tuple(a), make_tuple(b))
do_move = False
except InvalidMoveException as err:
print(err)
def createAllResearchTraining():
#Initialize empty lists to store all of the relevant information.
imageNames = []
CoordLeft = []
CoordRight = []
SpeciesList = []
NumFlowers = []
#with open('Research Map Data Association - Sheet1.csv', 'rb') as csvfile:
with open('EditedResearchMapData.csv', 'rb') as csvfile:
reader = csv.reader(csvfile, delimiter = ',')
i = 0
for row in reader:
if i == 0: #throw out the first row.
print(i)
elif len(row)<8:
print("Row too short")
else:
print(row)
if row[2] == '' or row[3] == '' or row[4] == '' or row[5] == '' or row[7] == '':
print('missing information')
else:
imageNames += [IMAGE_PATH + row[2] + '.jpg']
tupleLeft = make_tuple('(' + row[3] + ')')
CoordLeft += [tupleLeft]
tupleRight = make_tuple('(' + row[4] + ')')
CoordRight += [tupleRight]
SpeciesList += [row[5]]
NumFlowers += [float(row[7])]
i += 1
return imageNames, CoordLeft, CoordRight, SpeciesList, NumFlowers
def getTypeFromData(self,_data):
if self.variableType=="dynamic":
if _data.attrib.has_key('type'):
self.variableType = _data.attrib['type']
else: self.variableType = 'string'
if self.variableType=="string": return _data
if self.variableType=="int": return int(_data)
if self.variableType=="float": return float(_data)
if self.variableType=="bool": return (_data.lower() == 'true')
if self.variableType=="tuple": make_tuple(_data)
def onOk(self):
if DEBUG:
print >> sys.stderr, "values are:",
print >> sys.stderr, self.sp.get(),
print >> sys.stderr, self.ep.get(),
print >> sys.stderr, self.d.get(),
print >> sys.stderr, self.s.get()
sp = make_tuple(self.sp.get())
ep = make_tuple(self.ep.get())
d = int(self.d.get())
s = int(self.s.get())
self.top.destroy()
self.culebron.drag(sp, ep, d, s)
def __init__(self, config, communicator, defects, travel_time):
self.communicator = communicator
self.leaky_battery = ('True' == defects['leaky_battery'])
if self.leaky_battery:
print("The battery is set to leak")
self.uuid = config.get('uuid')
self.real_battery_size = config.getfloat('battery_size')
self.battery_size = self.real_battery_size
self.initial_location = Point(*make_tuple(config.get('c2_location')))
self.location = Point(*make_tuple(config.get('start_location')))
self.location_lock = asyncio.Lock()
self.start_time = 0
self.travel_time = travel_time
self.battery_id = 0
def main(config_file):
oldloop = asyncio.get_event_loop()
oldloop.close()
print("Bootstrapping drone configuration")
config = configparser.ConfigParser()
config.read(config_file)
num_drones = int(config["main"]["num_drones"])
df = config['detection']['data_folder']
if os.path.exists(df):
print("deleting data")
shutil.rmtree(df)
os.mkdir(df)
config = None
print("Generating subconfigurations")
configs = []
for i in range(0, num_drones):
config = configparser.ConfigParser()
config.read(config_file)
loc = tuple(
[float(x) for x in make_tuple(
config["telemetry"]["start_location"]
)]
)
nloc = (
loc[0] + (i * 0.0001 * (random.uniform(0, 2) - 1)),
loc[1] + (i * 0.0001 * (random.uniform(0, 2) - 1)),
loc[2]
)
config["telemetry"]["start_location"] = str(nloc)
configs.append(config)
return multi_drone_hybrid(configs)
def _recv_arrays(self):
"""Receive a list of NumPy arrays.
Parameters
----------
socket : :class:`zmq.Socket`
The socket to receive the arrays on.
Returns
-------
list
A list of :class:`numpy.ndarray` objects.
Raises
------
StopIteration
If the first JSON object received contains the key `stop`,
signifying that the server has finished a single epoch.
"""
headers = self.socket.recv_json()
if 'stop' in headers:
raise StopIteration
arrays = []
for header in headers:
data = self.socket.recv()
buf = buffer_(data)
array = np.frombuffer(buf, dtype=np.dtype(header['descr']))
array.shape = make_tuple(header['shape'])
if header['fortran_order']:
array.shape = header['shape'][::-1]
array = array.transpose()
arrays.append(array)
return arrays
def generate_data_set(args):
data_file = args.data
location_file = args.locations
locations = dict()
with open(location_file, 'r') as f:
for line in f:
addr, loc = line.strip().split(':')
locations[addr.strip()] = make_tuple(loc.strip())
uniq_set = set()
with open(data_file, 'r') as f:
lines = (line.strip() for line in f)
count = 0
for line in lines:
count += 1
if count == 1:
continue
info = parse_restaurant_info(line)
key = '%s, %s' % (info['addr'], info['zipcode'])
signature = '%s:%s:%s' % (info['name'], info['addr'], info['zipcode'])
if key in locations and signature not in uniq_set:
info['location'] = locations[key]
json_obj = json.dumps(info, ensure_ascii=False)
uniq_set.add(signature)
print json_obj
def results():
# prevent css caching
rand = random.randint(0,2500000)
c_cache = "../static/css/colors.css?" + str(rand)
cols = request.args.get('main_cols')
pallete = request.args.get('p_cols')
tups = make_tuple(cols)
tlist = []
hlist = []
for t in tups:
tlist.append(t[1])
hlist.append('%02x%02x%02x' % t[1])
primcol = tlist[0]
hcol = '%02x%02x%02x' % primcol
print("results pallete: %s" % pallete)
state = 'ran'
return render_template('results.html',
title='tagbar',
hashtag=request.args.get('tag'),
colors=cols,
primary=primcol,
hexcol=hcol,
hcs = hlist,
pcl=pallete,
dt=c_cache)
def get_patch_image(filename):
# we expect the filename format for be patch_(x, y, w, h).png"
a = "af"
dim_str = filename[filename.rindex('('):filename.rindex(')') + 1]
dim = make_tuple(dim_str)
data = matplotlib.image.imread(filename)[:, :, :3] # Use the last 3
return dim , data
def handle(self, *args, **options):
def _save_product(product_data):
# print(product_data)
product, created = Product.objects.get_or_create(
name=product_data[1],
description = product_data[2],
price = product_data[3],
discounted_price = product_data[4],
image = product_data[5],
image_2 = product_data[6],
thumbnail = product_data[7],
display = product_data[8]
)
print(product)
if created:
product.save()
return product
with open('tmp/data/products.json', 'r') as f:
data = f.readlines()
for row in data:
row_data = row.strip()
tuple_data = make_tuple(row_data)
try:
# print(tuple_data[0])
product = _save_product(tuple_data[0])
print("Product created :"+product)
except Exception:
print("Error")
#Save the product here
def main(nouns_loc, word2vec_loc, n_nouns, out_loc):
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s',
level=logging.INFO)
# Load trained Word2Vec model
model = Word2Vec.load(word2vec_loc)
logger.info('Word2Vec object loaded')
logger.info('Keeping %s nouns', n_nouns)
# Empty dictionary for noun to vector mapping
noun_to_vect_dict = {}
# Counter to know when to stop
counter = 0
with open(nouns_loc, 'r') as f:
while counter < int(n_nouns):
line = make_tuple(f.readline())
# Add noun and vector to mapping dictionary
noun = line[0]
noun_to_vect_dict[noun] = model[noun]
# Increment counter
counter += 1
logger.info('Pickling noun to vector dictionary')
# Pickle dictionary
with open(path.join(out_loc, 'noun_to_vect_dict_' + n_nouns + '.pkl'), 'w') as f:
pickle.dump(noun_to_vect_dict, f)
def filter_approx_distance(self, queryset, value):
""" Filters all results who's address object has a lat long approximatly value[0] from value[1]
"""
# Assume value is in the form (distance, lat, long)
try:
vals = make_tuple(value)
except:
# if something bad happened just fallabck to not working for now
return queryset
# remove queryset objects tha have no address
queryset = queryset.filter(address_object__isnull=False)
pi = 3.1415
f_lat = pi*(vals[1] - F('address_object__latitude'))/180.0
f_long = pi*(vals[2] - F('address_object__longitude'))/180.0
m_lat = 0.5*pi*(vals[1] + F('address_object__latitude'))/180.0
cosprox = 1 - (m_lat**2)/2.0 # approximate cosine
approx_dist = (6371**2)*(f_lat**2 + (cosprox*f_long)**2)
queryset = queryset.annotate(dist=(approx_dist - vals[0]**2)).annotate(flat=f_lat)
queryset = queryset.filter(dist__lte=0)
return queryset
def start_stream():
stop_event = threading.Event()
var_start_h.get()
var_start_m.get()
var_end_h.get()
var_end_m.get()
time_interval = (int(var_start_h.get()), int(var_start_m.get()), int(var_end_h.get()), int(var_end_m.get()))
writeToFile("time_schedule.in", time_interval)
if not check_times(time_interval):
print time_interval
print "Error in data_scheduler, wrong format in the file."
return
if start_interval_reached(time_interval[0], time_interval[1]):
time = calculate_time(time_interval[0], time_interval[1])
print time
threading.Timer(time, get_continuously_data, [make_tuple(str(time_interval))]).start()
while not stop_event.is_set():
if (end_interval_reached1(time_interval[2], time_interval[3], 3)):
show_piechart_now(time_interval)
stop_event.set()
def make_reservation(self):
reservation_name = input("Choose name: ")
number_of_tickets = input("Choose number of tickets: ")
self.show_movies()
while True:
reservation_movie_id = input("Choose movie id: ")
self.__cinema.get_num_of_free_seats_by_movie_id(
reservation_movie_id)
wanted_projection_id = input("Choose projection id: ")
if self.how_many_free_seats(reservation_movie_id, number_of_tickets):
break
else:
print("There are no more available seats! Enter new id: ")
list_of_not_available_seats = self.__cinema.show_all_available_spots_matrix(
wanted_projection_id)
matrix = self.matrix_print(list_of_not_available_seats)
count = 0
list_of_reserved_seats = []
while int(count) < int(number_of_tickets):
seat_tuple_str = input("Choose a seat: ")
seat_tuple = make_tuple(seat_tuple_str)
if int(seat_tuple[0]) > 10 or int(seat_tuple[0]) < 1 or int(seat_tuple[1]) > 10 or int(seat_tuple[1]) < 1:
print("There is no such seat")
elif matrix[int(seat_tuple[0]) - 1][int(seat_tuple[1]) - 1] == 'X':
print("This seat is taken")
else:
count += 1
list_of_reserved_seats.append(seat_tuple)
res = {}
res["res_name"] = reservation_name
res["list_of_seats"] = list_of_reserved_seats
res["projection_id"] = wanted_projection_id
print("If you want to save your reservation type finalize")
return res
def recipe_read(recipe: dict) -> tuple:
bright, ttime, pause_time = recipe
if type(bright) is str and bright[0:3] in 'rnd':
val = make_tuple(bright[3:])
val = (val[0], val[1]) if (val[0] < val[1]) else (val[1], val[0])
bright = random.randint(*val)
return bright, ttime, pause_time
def test_basics(self):
# Setup the files with expected content.
temp_folder = tempfile.mkdtemp()
self.create_file(os.path.join(temp_folder, 'input.txt'), FILE_CONTENTS)
# Run pipeline
# Avoid dependency on SciPy
scipy_mock = MagicMock()
result_mock = MagicMock(x=np.ones(3))
scipy_mock.optimize.minimize = MagicMock(return_value=result_mock)
modules = {
'scipy': scipy_mock,
'scipy.optimize': scipy_mock.optimize
}
with patch.dict('sys.modules', modules):
from apache_beam.examples.complete import distribopt
distribopt.run([
'--input=%s/input.txt' % temp_folder,
'--output', os.path.join(temp_folder, 'result')])
# Load result file and compare.
with open_shards(os.path.join(temp_folder, 'result-*-of-*')) as result_file:
lines = result_file.readlines()
# Only 1 result
self.assertEqual(len(lines), 1)
# parse result line and verify optimum
optimum = make_tuple(lines[0])
self.assertAlmostEqual(optimum['cost'], 454.39597, places=3)
self.assertDictEqual(optimum['mapping'], EXPECTED_MAPPING)
production = optimum['production']
for plant in ['A', 'B', 'C']:
np.testing.assert_almost_equal(production[plant], np.ones(3))
def open_model_file(file_name):
with open(file_name) as data_file:
model = json.load(data_file)
print 'read model!'
model = {make_tuple(str(key)): value for key, value in model.iteritems()}
print 'convert model!'
return model
def getJourneys(filename, algorithm): """ @param filename: name of text file with the following format (space separated) - one journey per line <lat> <long> <startTime> @param algorithm: should given, start and end node and time, returns a journey @return: list of journeys startTime should be time in same format as time in crime data """ journeys = [] with open(filename) as f: lines = f.readlines() for line in lines: triple = line.split() journey = algorithm(make_tuple(triple[0]), make_tuple(triple[1]), triple[2]) journeys.append(journey) return journeys
def get_sorted_average_data(file_name, inflation_rates, compute_function):
f = open(file_name, 'r')
raw_lines = f.readlines()
tuple_lines = [make_tuple(line.strip()) for line in raw_lines]
computed_data = list(compute_function(tuple_lines, MAX_GROWTH, inflation_rates))
averaged_data = list(average_data(computed_data))
sorted_average_data = sorted(averaged_data, key=lambda t: (int(t[0].split("-")[0]), int(t[0].split("-")[1])))
return sorted_average_data
def to_tuple(inval):
try:
val = make_tuple(inval)
except (ValueError, SyntaxError):
raise ValidationError("Invalid input: '{0}'".format(inval))
if not isinstance(val, collections.Iterable):
raise ValidationError("Value must be a tuple")
return val
def ParseFromString(cmdContent):
#invoke intelliSense
strRep = str(cmdContent)
segments = strRep.split(':')
cmd = str(segments[0])
instanceId = int(segments[1])
requestNumber = int(segments[2])
value = make_tuple(segments[3])
return CommandObject(cmd,requestNumber,instanceId,value)
def __init__(self,photo_config,config):
super(Picture, self).__init__()
self.photo_config = photo_config
self.config = config
self.source = photo_config.path
self.bind(source=self.set_source)
self.pos = make_tuple(photo_config.pos)
self.rotation = float(photo_config.rotation)
self.scale = float(photo_config.scale)
def callback(*args):
new_val = str_v.get()
if isinstance(initial, tuple):
new_val = make_tuple(new_val)
elif isinstance(initial, int):
new_val = int(new_val)
set_conf_parameter(name, new_val)
def set_motors_registers(motors_register_value):
""" Allow lot of motors register settings with a single http request
Be carefull: with lot of motors, it could overlap the GET max
lentgh of your web browser
"""
for m_settings in motors_register_value.split(";"):
settings = m_settings.split(":")
rr.set_motor_register_value(settings[0], settings[1], make_tuple(settings[2]))
return "Done!"
def to_tuple(inval):
if isinstance(inval, list):
return tuple(inval)
try:
val = make_tuple(inval)
except (ValueError, SyntaxError):
raise ValidationError('Invalid input: "{0}"'.format(inval))
if not isinstance(val, collections.Iterable):
raise ValidationError('Value must be a tuple')
return val
def to_json(request):
# try to parse json
try:
return json.loads(request)
except ValueError:
try:
return make_tuple(request)
except ValueError:
return None
return None
def onOk(self, event=None):
if DEBUG:
print >> sys.stderr, "onOK()"
print >> sys.stderr, "values are:",
print >> sys.stderr, self.sp.get(),
print >> sys.stderr, self.ep.get(),
print >> sys.stderr, self.d.get(),
print >> sys.stderr, self.s.get(),
print >> sys.stderr, self.units.get()
sp = make_tuple(self.sp.get())
ep = make_tuple(self.ep.get())
d = int(self.d.get())
s = int(self.s.get())
self.cleanUp()
# put focus back to the parent window's canvas
self.culebron.canvas.focus_set()
self.destroy()
self.culebron.drag(sp, ep, d, s, self.units.get())
def parse_lat_lon(self, s):
"""
Read the latitude/longitude row into a tuple
and return lat/lon values
"""
if s:
t = make_tuple(s)
return t[1], t[0]
else:
return (None, None)
# Filter the interactions dataframe to these interactions only
interactions_filtered = interactions.loc[
interactions.prot_A.isin(proteins_covered)
& interactions.prot_B.isin(proteins_covered), ]
print('{} interactions are covered from the {} proteins above'.format(
interactions_filtered.shape[0], len(proteins_covered)))
# Write the final set of interactions to file
interactions_filtered_tsv = pathlib.Path.joinpath(
output_dir, 'interactions_filtered.tsv')
interactions_filtered.to_csv(interactions_filtered_tsv,
sep='\t',
index=False)
# Also provide a tsv file that contains the uniprot-uniprot interaction
# as ncbi-ncbi interaction (for the negatives set construction)
ncbi_interactions_tsv = pathlib.Path.joinpath(output_dir,
'ncbi_interactions.tsv')
# This is a list of strings ["('intA', 'intB')" , ...]
interaction_pairs = interactions_filtered['interaction'].to_list()
with open(ncbi_interactions_tsv, 'w') as fout:
for pair in interaction_pairs:
# Convert the literal tuple string to actual tuple
# TO DO
# Just make it a tuple to begin with in the interactions file...
p = make_tuple(pair)
fout.write('{}\t{}\t{}\t{}\n'.format(
p[0], p[1], uniprot2ncbi_protein_mapping.get(p[0]),
uniprot2ncbi_protein_mapping.get(p[1])))
print('Done!')
def main():
global connect, cursor
id_query = []
SQL = '''SELECT * FROM omsk.noise_to_air ORDER BY id ASC;'''
cursor.execute(SQL)
last_ID_list = cursor.fetchall()
SQL = '''SELECT (time_track) FROM omsk.tracks ORDER BY time_track desc LIMIT 1;'''
cursor.execute(SQL)
last_time_air = cursor.fetchall()
# print('LAST TIME TRACK', last_time_air)
if len(last_time_air) > 0: # the table can be empty..
last_time_air = last_time_air[0][0]
we_have_data = False
for ID in last_ID_list:
## separate query to not have a headache with formatiing.. since string+datetime..
SQL = '''SELECT (base_name) FROM omsk.noise WHERE id = (%s)'''
data = ID
cursor.execute(SQL, data)
base_name = cursor.fetchall()[0][0]
SQL = '''SELECT (time_noise) FROM omsk.noise WHERE id = (%s)'''
data = ID
cursor.execute(SQL, data)
time_noise = cursor.fetchall()[0][0]
# print('BASE NAME', base_name)
# print('TIME NOISE', time_noise)
if last_time_air - time_noise >= minimum_time: #minimum_time: # if noise_time is greater than noise datetime by 10 sec -> fine
# if base_name == VNK001 -> use distance_1, VNK002 -> distance_2
# print('10 SEC PASSED!')
SQL = {
'OMSK001':
'''SELECT (track, distance_1, time_track) FROM omsk.tracks WHERE time_track >= (%s)::timestamp - INTERVAL '10 seconds' and time_track <= (%s)::timestamp + INTERVAL '10 seconds' and distance_1 IS NOT NULL ORDER BY distance_1 asc LIMIT 1;''',
'OMSK002':
'''SELECT (track, distance_2, time_track) FROM omsk.tracks WHERE time_track >= (%s)::timestamp - INTERVAL '10 seconds' and time_track <= (%s)::timestamp + INTERVAL '10 seconds' and distance_2 IS NOT NULL ORDER BY distance_2 asc LIMIT 1;''',
'OMSK003':
'''SELECT (track, distance_3, time_track) FROM omsk.tracks WHERE time_track >= (%s)::timestamp - INTERVAL '10 seconds' and time_track <= (%s)::timestamp + INTERVAL '10 seconds' and distance_3 IS NOT NULL ORDER BY distance_3 asc LIMIT 1;''',
'OMSK004':
'''SELECT (track, distance_4, time_track) FROM omsk.tracks WHERE time_track >= (%s)::timestamp - INTERVAL '10 seconds' and time_track <= (%s)::timestamp + INTERVAL '10 seconds' and distance_4 IS NOT NULL ORDER BY distance_4 asc LIMIT 1;'''
}
SQL = SQL[base_name]
data = (time_noise, time_noise)
cursor.execute(SQL, data)
track_distance = cursor.fetchall()
# print('TRACK DISTANCE', track_distance)
we_have_data = True
if len(track_distance) == 0: ## no aircraft within 10 sec..
SQL = '''
DELETE FROM omsk.noise_to_air WHERE id = (%s);
'''
data = ID
# print('NO aircraft within 10s')
else:
SQL = '''
UPDATE omsk.noise SET (track, distance, aircraft_time) = (%s, %s, %s) WHERE id=(%s);
DELETE FROM omsk.noise_to_air WHERE id = (%s);
'''
data = make_tuple(track_distance[0][0]) + ID + ID
# print('update noise ID: ', ID)
id_query.append(cursor.mogrify(SQL, data).decode('utf-8'))
if we_have_data:
full_query_from_ID_list = ''.join([x for x in id_query])
cursor.execute(full_query_from_ID_list)
connect.commit()
def _read_class_file(self, class_file):
data_base_path = os.path.dirname(__file__)
data_file = os.path.join(data_base_path, class_file)
# base_path = os.path.realpath(os.path.join(self.data_dir))
colours = [make_tuple(line.rstrip()) for line in open(data_file)]
return colours
def train(self):
config = ConfigParser.RawConfigParser()
config.read(self.conf_path)
inputpath = config.get('query_correction', 'training_set_path')
training_set = []
with open(inputpath) as f:
for line in f:
training_set.append(make_tuple(line))
lbda = 10
mu1 = 1000
mu2 = 1000
mu3 = 1000
for entry in training_set:
wrong_elem = entry[0]
correct_elem = entry[1]
wrong_list = wrong_elem.split(" ")
for x in range(1, len(wrong_list)):
corrected_output = self.api_for_training(wrong_list[:x])
correct_flag = True
if len(corrected_output) == len(correct_elem):
for idx in range(0, len(corrected_output)):
if not corrected_output[idx] == correct_elem[idx]:
correct_flag = False
break
else:
correct_flag = False
if not correct_flag: #query generation does not equal to training datapoint
for idx in range(
0,
min(len(corrected_output), len(correct_elem),
len(wrong_list))):
current_query = wrong_list[idx]
corrected_query = corrected_output[idx]
corrected_query_type = self.get_error_type(
current_query, corrected_query)
correct_word = correct_elem[idx]
correct_type = self.get_error_type(
current_query, correct_word)
corrected_prev_word = None
corrected_prev_type = None
correct_prev_word = None
correct_prev_type = None
corrected_skip_word = None
correct_skip_word = None
if idx >= 1:
corrected_prev_word = corrected_output[idx - 1]
corrected_prev_type = self.get_error_type(
wrong_list[idx - 1], corrected_prev_word)
correct_prev_word = correct_elem[idx - 1]
correct_type = self.get_error_type(
wrong_list[idx - 1], correct_prev_word)
if idx >= 2:
corrected_skip_word = corrected_output[idx - 2]
correct_skip_word = correct_elem[idx - 2]
lbda+=self.score_obj.phi_func(correct_skip_word,correct_prev_word,correct_prev_type,correct_word,correct_type)-\
self.score_obj.phi_func(corrected_skip_word,corrected_prev_word,corrected_prev_type,corrected_query,corrected_query_type)
mu1 += self.score_obj.f1_func(
correct_word, correct_type,
current_query) - self.score_obj.f1_func(
corrected_query, corrected_query_type,
current_query)
mu2 += self.score_obj.f2_func(
correct_word, correct_type,
current_query) - self.score_obj.f2_func(
corrected_query, corrected_query_type,
current_query)
mu3 += self.score_obj.f3_func(
correct_word, correct_type,
current_query) - self.score_obj.f3_func(
corrected_query, corrected_query_type,
current_query)
lbda /= len(wrong_list)
mu1 /= len(wrong_list)
mu2 /= len(wrong_list)
mu3 /= len(wrong_list)
lbda /= len(training_set)
mu1 /= len(training_set)
mu2 /= len(training_set)
mu3 /= len(training_set)
output = {}
output["_lbda"] = lbda
output["_mu1"] = mu1
output["_mu2"] = mu2
output["_mu3"] = mu3
self.score_obj.write_params(self.conf_path, output)
return output
'nan', 'nan', 'nan', 'nan', 'nan', 'nan',
'nan', 'nan', 'nan', 'nan', 'nan', 'nan',
'nan', 'nan', 'nan', 'nan', 'nan', 'nan',
'nan', 'nan', 'nan', 'nan', 'nan', 'nan',
'nan', 'nan', 'nan', 'nan', 'nan', 'nan',
'nan', 'nan', 'nan', 'nan', 'nan', 'nan',
'nan', 'nan', 'nan', 'nan'
]
patient.append(nan)
# patient = pd.DataFrame(patient)
continue
matches = 0
for nose_point in nose_points:
if not pd.isnull(nose_point):
nose_point = make_tuple(nose_point)
nose_x = nose_point[0]
nose_y = nose_point[1]
if (x_min <= nose_x <= x_max) & (y_min <= nose_y <=
y_max):
matches += 1
for reye_point in reye_points:
if not pd.isnull(reye_point):
reye_point = make_tuple(reye_point)
reye_x = reye_point[0]
reye_y = reye_point[1]
if (x_min <= reye_x <= x_max) & (y_min <= reye_y <=
y_max):
matches += 1
def unstring_keys(movie_data):
positions = movie_data["positions"]
new_positions = {make_tuple(k): v for k, v in positions.items()}
#print("new", new_positions.keys())
movie_data["positions"] = new_positions
return movie_data
import collections
import sys
from utils import *
import re
from ast import literal_eval as make_tuple
from timeit import default_timer as timer
for line in sys.stdin.readlines():
depth = 0
state = False
pathHolder = []
line.replace(" ", "")
pathS = ""
line = make_tuple(line)
#line = re.sub('[,]', '', line)
current = ()
for x in range(len(line)):
temp = (line[x])
current = current + (line[x], )
stack = []
path = []
pathQ = []
if check_clean(current):
#print "Done cleaning"
break
else:
start = timer()
stack.append(current)
path.append(current)
pathQ.append(pathS)
ret = (DLS_beta_pure(5, stack, path, pathQ))
print ret
cw = createCrossword(size=cwSize, sortedData=sortedData, blanks=grid) print 'Creating CSP' csp = createCrosswordCSP(cw) start = datetime.now() print 'In Backtrack Search' search = BacktrackingSearch() solution = search.solve_test(weights, cw, csp, mcv=True, ac3=True) end = datetime.now() addAssignmentsToGrid(cw, solution) print 'Computing Accuracy' numWordsAssigned = 0 for var in solution.keys(): if len(make_tuple(var)) > 2: numWordsAssigned += 1 solvedPerc = numWordsAssigned*1.0/len(cw.words.keys()) print solution print cw.grid totalTime += (end-start) totalAccuracy += solvedPerc if len(solution.keys()) == csp.numVars: numSolved += 1 solved = True # For logging experimental results df = open(dataFileName, 'a')
return 1
return 1
# plot_points(DIR_NAME+'train_file.txt', 'blue', 'red')
# plot_points(DIR_NAME+'test_file.txt', 'red', 'green')
# plot_line(get_final_weights(), 'red')
# plot_line(get_ideal_weights(), 'blue')
EXPERIMENT_RESULTS = []
m = 0
for rule in get_updating_rules():
m += 1
x = make_tuple(rule)
sample = x[1]
# if sample[2] == 0:
# plt.scatter([sample[0]], [sample[1]], color='red', s=20)
# else:
# plt.scatter([sample[0]], [sample[1]], color='green', s=20)
wt = x[0]
# l = plot_line(wt, 'black')
# plot_line(get_ideal_weights(), 'blue')
# plot_points(DIR_NAME+'train_file.txt', 'blue', 'red', alpha=0.3)
arc = calculate_incorrect_arc_ratio(wt)
# plt.text(0, 1, 'Incorrect Arc Ratio: %s' % arc,
# horizontalalignment='left',
# verticalalignment='top',
# transform=plt.gca().transAxes)
EXPERIMENT_RESULTS.append((m, arc))
def getBatchParams(self):
errors = []
waves = []
innerNAs = []
outerNAs = []
bstups = []
wave = self.batch_wave.text()
if not any(wave.split(",")):
errors.append("Must include at least one wavelength")
else:
for w in wave.split(","):
if not w:
continue
elif not w.isdigit() or not (300 < int(w) < 800):
errors.append(
'Wavelength "{}" not an integeger from 300-800'.format(
w))
else:
waves.append(float(w) / 1000)
innerNA = self.batch_innerNA.text()
if not any(innerNA.split(",")):
errors.append("Must include at least one innerNA")
else:
for w in innerNA.split(","):
try:
w = float(w)
if not (0 < w < 0.7):
raise ValueError("")
else:
innerNAs.append(w)
except ValueError:
errors.append(
'InnerNA "{}" not a float from 0-0.7'.format(w))
outerNA = self.batch_outerNA.text()
if not any(outerNA.split(",")):
errors.append("Must include at least one outerNA")
else:
for w in outerNA.split(","):
try:
w = float(w)
if not (0 < w < 0.7):
raise ValueError("")
else:
outerNAs.append(w)
except ValueError:
errors.append(
'outerNA "{}" not a float from 0-0.7'.format(w))
beamSpacing = self.batch_beamSpacing.text()
tupstring = "(\\(.*?\\))"
tups = re.findall(tupstring, beamSpacing)
if not len(tups):
errors.append('Must include at least one "(beams, spacing)" tuple')
else:
from ast import literal_eval as make_tuple
for tup in tups:
try:
t = list(make_tuple(tup))
if len(t) != 2:
errors.append(
"(beams, spacing) length not equal to 2: {}".
format(t))
else:
if not (1 <= t[0] <= 100):
errors.append(
'Number of Beams "{}" not an int between 0-100'
.format(t[0]))
t[0] = None
else:
t[0] = int(t[0])
if not (0 <= t[1] <= 50):
errors.append(
'Beam Spacing "{}" not a float between 0-50'.
format(t[1]))
t[1] = None
else:
t[1] = float(t[1])
if all([x is not None for x in t]):
bstups.append(tuple(t))
except ValueError:
t = tup.strip("(").strip(")").split(",")
if len(t) != 2:
errors.append(
"(beams, spacing) length not equal to 2: {}".
format(t))
else:
if t[0].lower().strip() == "fill":
t[0] = "fill"
elif t[0].isdigit() and (1 <= int(t[0]) <= 100):
t[0] = int(t[0])
else:
errors.append(
'Number of Beams "{}" not an int between 0-100 or keyword "fill"'
.format(t[0]))
t[0] = None
if t[1].lower().strip() == "auto":
t[1] = "auto"
elif str_is_float(t[1]) and (0 < float(t[1]) <= 50):
t[1] = float(t[1])
else:
errors.append(
'Beam Spacing "{}" must either be float between 0-50 or keyword "auto"'
.format(t[1]))
t[1] = None
if all([t[0], t[1]]):
t[1] = None if t[1] == "auto" else t[1]
bstups.append(tuple(t))
xshift = self.batch_xShift.text()
if not any(xshift.split(":")):
xshifts = [0]
else:
try:
a = [float(x) for x in xshift.split(":")]
if len(a) == 1:
xshifts = a
else:
a[1] += 0.000001 # include stop index
xshifts = np.arange(*a)
xshifts = sorted(
list(
set([
round(y, 2) for y in xshifts if -100 < y < 100
])))
except TypeError as e:
errors.append("X Shift Range not valid: {}".format(e))
yshift = self.batch_yShift.text()
if not any(yshift.split(":")):
yshifts = [0]
else:
try:
a = [float(y) for y in yshift.split(":")]
if len(a) == 1:
yshifts = a
else:
a[1] += 0.000001 # include stop index
yshifts = np.arange(*a)
yshifts = sorted(
list(
set([
round(y, 2) for y in yshifts if -100 < y < 100
])))
except TypeError as e:
errors.append("Y Shift Range not valid: {}".format(e))
tilt = self.batch_tilt.text()
if not any(tilt.split("-")):
tilts = [0]
else:
try:
a = [float(ti) for ti in tilt.split(":")]
if len(a) == 1:
tilts = a
else:
a[1] += 0.000001 # include stop index
tilts = np.arange(*a)
tilts = sorted(
list(
set([
round(ti, 2) for ti in tilts
if -1.5 <= ti <= 1.5
])))
except TypeError as e:
errors.append("Tilt Range not valid: {}".format(e))
# yshift = self.batch_yShift.text()
# tilt = self.batch_tilt.text()
if len(errors):
self.show_error_window(
"There were some errors in the batch slm settings:",
title="Batch SLM Error",
info="\n".join(errors),
)
else:
from itertools import product
a = [waves, innerNAs, outerNAs, bstups, xshifts, yshifts, tilts]
combos = [list(c) for c in product(*a) if c[1] < c[2]]
for c in combos:
# if only 1 beam, force tilt to 0
if isinstance(c[3][0], int) and c[3][0] == 1:
c[6] = 0
combos = list(set([tuple(c) for c in combos
])) # get rid of duplicates, like from tilt=0
if not len(combos):
raise InvalidSettingsError(
"No valid combinations!",
"Is there at least one outerNA that is greater than the min innerNA?",
)
return combos
# "(8, 11)":4,
# "(0, 0)":1,
# "(18, 11)":5,
# "(18, 13)":1,
# "(13, 13)":7,
# "(13, 11)":2,
# "(14, 8)":1,
# "(10, 10)":3
# }
ls = []
for s in range(0, 19):
ls.append([0 for s in range(0, 19)])
for d in data:
tup = make_tuple(d)
if tup[0] == None or tup[1] == None:
continue
ls[tup[1]][tup[0]] = data[d]
# import pprint; pprint.pprint(ls)
print "labels", '\t',
for s in scores:
print str(s) + 'o', '\t',
print '\n'
for s in scores[::-1]:
print str(s) + 'i', '\t',
for t in scores:
print ls[t][s], '\t',
import collections
import sys
from utils import *
import re
from ast import literal_eval as make_tuple
import utils
from timeit import default_timer as timer
current = []
for file in sys.stdin.readlines():
line = make_tuple(file)
#print line
row = []
for x in line:
row.append(x)
current.append(row)
#print(current)
#print(len(current))
"""
for x in range(len(current)-1):
for y in range(len(current)-1):
if current[x][y] != 0 and current[x][y] != 1:
print "not valid"
quit()
if current[len(current)-1][0] < 0 or current[len(current)-1][0] > len(current)-2:
print False
quit()
elif current[len(current)-1][1] < 0 or current[len(current)-1][1] > len(current)-2:
print False
quit()
def set_table_from_json_str(self, json_str):
tmp_obj = json.loads(json_str)
self.table = dict([(make_tuple(key), val)
for key, val in tmp_obj.iteritems()])
def get_random_values(distribution: dict, n=1):
"""Receives a `distribution` and outputs `n` random values
Distribution format: { \'name\': str, \'parameters\': tuple }"""
dist = getattr(scipy.stats, distribution['name'])
param = make_tuple(distribution['parameters'])
return dist.rvs(*param[:-2], loc=param[-2], scale=param[-1], size=n)
if __name__ == '__main__':
#
import sys, os
sys.path.append(os.path.abspath('../layer'))
from anchor_box_layer import *
from focal_loss_layer import *
from rpn_focal_loss_layer import *
from merge_rpn_cls_layer import *
from yolo_target_layer import *
from smoothed_focal_loss_layer import *
from roi_transform_layer import *
from iou_loss_layer import *
args = parse_args()
if not args.epoch:
args.epoch = 1
if not args.data_shape:
args.data_shape = 416
if not args.label_shape:
args.label_shape = (1, 50, 6)
else:
args.label_shape = make_tuple(args.label_shape)
try:
net, _, _ = mx.model.load_checkpoint(args.prefix, args.epoch)
except:
net, _, _ = mx.model.load_checkpoint(args.prefix, 0)
estimate_mac(net, args.data_shape, args.label_shape)
# mx.viz.print_summary(net, {'data': (1, 3, 384, 384), 'label': (1, 50, 6)})
def get_corner_coords(imagefilename,
backend=params['backend'],
size=make_tuple(params['pattern_size']),
show_figure=False,
save_figure=params['output_img_with_dectected_corners']):
if backend == "matlab":
try:
import matlab.engine
print "Matlab is used as backend for detecting corners"
except ImportError:
print "matlab.engine can not be found!"
print "To use detectCheckerboardPoints function of matlab in python, matlab.engine for python should be installed!"
sys.exit(0)
eng = matlab.engine.start_matlab()
imagePoints, boardSize, imagesUsed = eng.detectCheckerboardPoints(
imagefilename, nargout=3)
print boardSize, imagesUsed
if not imagesUsed:
print "Corners can not be detected!"
return None
np_imagePoints = np.array(imagePoints)
if save_figure or show_figure:
img = cv2.imread(imagefilename)
size = tuple((np.array(boardSize).astype(np.int32) - 1).flatten())
cv2.drawChessboardCorners(img, size,
np_imagePoints.astype(np.float32), 1)
if save_figure:
save_imagefilename = "output/img_corners/" + (
imagefilename.split("/")[-1]).split(".")[
0] + "_detected_corners" + "." + params['image_format']
cv2.imwrite(save_imagefilename, img)
print "Image with detected_corners is saved in " + save_imagefilename
if show_figure:
cv2.imshow("image with detected corners", img)
while True:
k = cv2.waitKey(1)
if k == 27:
cv2.destroyAllWindows()
break
return np_imagePoints
elif backend == "opencv":
print "OpenCV" + str(
cv2.__version__) + " is used as backend for detecting corners"
img = cv2.imread(imagefilename)
print img.shape
ret, corners = cv2.findChessboardCorners(
img,
size,
flags=cv2.cv.CV_CALIB_CB_ADAPTIVE_THRESH +
cv2.cv.CV_CALIB_CB_NORMALIZE_IMAGE + cv2.CALIB_CB_FAST_CHECK)
if not ret:
print "Corners can not be detected!"
return None
cv2.drawChessboardCorners(img, size, corners, ret)
cv2.namedWindow("img", cv2.WINDOW_NORMAL)
cv2.imshow('img', img)
while True:
k = cv2.waitKey(1)
if k == 27:
cv2.destroyAllWindows()
break
save_imagefilename = "output/img_corners/" + (
imagefilename.split("/")[-1]
).split(".")[0] + "_detected_corners" + "." + params['image_format']
cv2.imwrite(save_imagefilename, img)
return corners
else:
AssertionError("Please input the right backend for corner detection")
import pandas as pd
import csv
from ast import literal_eval as make_tuple
df = pd.read_csv('entities.csv')
with open('new_entities.csv', mode='w', encoding='utf8') as csv_file:
fieldnames = ['phrase', 'entity', 'count']
writer = csv.DictWriter(csv_file, fieldnames=fieldnames)
writer.writeheader()
for index, row in df.iterrows():
curr_tuple = make_tuple(row['entity'])
writer.writerow({
'phrase': curr_tuple[0],
'entity': curr_tuple[1],
'count': row['count']
})
if not added:
accumulator_times.append(new_times)
accumulator_lattices.append(new_lattices)
return (replicate,(accumulator_times,accumulator_lattices))
if __name__ == "__main__":
if len(sys.argv) < 7:
print "Usage: output_dir output_indices hdfs_sfile min_replicate max_replicate species+ [--local] [--interactive]"
quit()
outputDir = sys.argv[1]
outputIndices = make_tuple(sys.argv[2])
if type(outputIndices) != type((0,)):
print "Error: output indices must be specified in as a python formatted tuple, got: %s"%(type(outputIndices))
quit()
filename = sys.argv[3]
replicates = range(int(sys.argv[4]),int(sys.argv[5])+1)
species = []
interactive = False
local = False
for i in range(6,len(sys.argv)):
if sys.argv[i] == "--interactive":
interactive=True
continue
if sys.argv[i] == "--local":
local=True
node = node.parent
def search(self):
for i in range(1000):
node = self.selection()
node = self.expansion(node)
reward = self.rollout(node)
self.backpropagation(node, reward)
return self.root.get_best_child()
if __name__ == "__main__":
state = FiveInRowState(size=5)
print(state)
while True:
move = make_tuple(input("Enter your move in the form (row, column): "))
state = state.make_move(move)
print(state)
root = Node(state)
search = MonteCarloTreeSearch(root)
child = search.search()
state = child.state
print(child.state)
# state = child.state.make_move((3,4))
# root = Node(state)
# search = MonteCarloTreeSearch(root)
# child = search.search()
# print(child.state)
# state = child.state.make_move((3, 5))
# root = Node(state)
# search = MonteCarloTreeSearch(root)
parser.add_argument("--schan_vals", help="schan parameter search space", type=str, default='(1,30,1)')
args = parser.parse_args()
# get the data set name
print(args.validation_data_path)
dataset_name = args.validation_data_path.split('/')
if dataset_name[-1]=='':
dataset_name = dataset_name[-3]
else:
dataset_name = dataset_name[-2]
print(dataset_name)
print('Evaluation on {} data set'.format(dataset_name))
# make tuples
args.sxy_vals = make_tuple(args.sxy_vals)
args.schan_vals = make_tuple(args.schan_vals)
args.compat_vals = make_tuple(args.compat_vals)
# tune parameters
compat_best, sxy_best, srgb_best, dice_coefficient, dice_coefficients = crf_parameter_tuning(args.validation_data_path,
args.model_filename,
args.compat_vals,
args.sxy_vals,
args.schan_vals,
args.pairwise_feature,
args.image_preprocessing)
# write best configurations on disc
if not path.exists(args.output_path):
makedirs(args.output_path)
with open(path.join(args.output_path, dataset_name + '_' + args.pairwise_feature + '_crf_optimization.txt'), 'w') as file:
import csv
import collections
import pprint
from ast import literal_eval as make_tuple
type_chain = {'PS': ['NP'], 'NP': ['HF'], 'HF': ['4G', '4GF']}
fate_classes_diffs = collections.defaultdict(dict)
fate_classes_starts = collections.defaultdict(dict)
with open('results/norm/actions_norm.csv', 'r') as f:
reader = csv.reader(f)
for row in reader:
start_state, start_type = make_tuple(row[0])
end_state, end_type = make_tuple(row[-2])
next_types = type_chain[start_type]
for ele in row:
ele_state, ele_type = make_tuple(ele)
if ele_type in next_types:
diffs = [
ele_state[i] - start_state[i]
for i in range(len(ele_state))
]
if start_type in fate_classes_diffs[end_type]:
fate_classes_diffs[end_type][start_type].append(diffs)
else:
fate_classes_diffs[end_type][start_type] = [diffs]
break
with open('results/norm/fate_transitions_norm.csv', 'w') as f:
writer = csv.writer(f)
def input_to_tuple(str):
return make_tuple("(" + str + ")")
def process_academic_lecturer():
try:
# check request method
# if request method = POST
if request.method == 'GET':
if request.args.get('id'):
return jsonify(
get_academicLecturer_byID(request.args.get('id')))
# call get all academic lecturer method from academic module
return jsonify(get_all_academic_lecturer())
# if request method = POST
elif request.method == 'POST':
# get data from json request
course_id = request.json['course_id']
course_class = request.json['course_class']
lecturer = request.json['lecturer']
total_credit = request.json['total_credit']
ret = []
for lecturer in lecturer:
data = make_tuple(lecturer)
# build new academic lecturer object with initial value
new_academic_lecturer = academic_lecturer(
course_id=course_id,
course_class=course_class,
total_credit=total_credit,
lecturer_nip=data[0],
lecturer_credit=data[1])
# call the save method from academic lecturer object
ret.append(new_academic_lecturer.save())
return jsonify(ret)
# if request method = PUT
elif request.method == 'PUT':
# get id from query
id = request.args.get('id')
# build new academic lecturer object
academic_lecturer_object = academic_lecturer()
# call the edit method from academic lecturer object
ret = academic_lecturer_object.edit(id, request.json)
return jsonify(ret)
# if request method = DELETE
elif request.method == 'DELETE':
# get id from query
id = request.args.get('id')
# build new announcement object
academic_lecturer_object = academic_lecturer()
# call the delete method from announcement object
ret = academic_lecturer_object.delete(id)
return jsonify(ret)
# if request method is unknown
else:
return jsonify({
'status':
500,
'message':
'Sorry, your request method is not recognized!'
})
except Exception as e:
res = {'message': e.args}
return res
def plotHashtags(filename, savePng=True):
"""Created a stacked bar plot for the given hashtag data csv
:input: filename: path to file of csv
savePng: if user wants the plot saved as a png
:type: None
:return: None
:type: None
"""
# error checking
assert isinstance(filename, basestring)
try:
df = pd.read_csv(filename)
except:
print "error reading file " + filename
return
col_number = df.shape[1]
row_number = df.shape[0]
#print hasthag counts to console
# for i in df.ix[y]:
# print i
data = [[] for i in range(col_number)]
col_number2=str(col_number-2)
temp=[]
for row in df.iterrows():
index, data = row
temp.append(data.tolist())
xDates = df['date']
data=[]
data_numer=[]
data_hashtag=[]
# get data except dates
for i in df.loc[:,col_number2:'0']:
data.append( df[i].tolist())
# extract tuple data
for i in data:
x=[]
y=[]
for j in i:
x.append(make_tuple(j)[1])
y.append(make_tuple(j)[0])
data_numer.append(x)
data_hashtag.append(y)
df=pd.DataFrame(data_numer)
df=df.transpose()
# plot counts
ax = df.plot.bar(stacked=True,legend=False);
#label totals at top of bar
for i in ax.patches[len(ax.patches)-row_number:]:
plt.text(i.get_x(),i.get_y()+75, int(i.get_y()),fontsize=10)
plt.gcf().subplots_adjust(bottom=0.15)
plt.xticks(range(10),xDates)
plt.xticks(rotation=30)
plt.ylabel('number of hashtags')
plt.xlabel('date')
height = 2000
for i in temp:
plt.text(10,height,i)
height = height-200
# if set, save as png
if savePng:
plt.savefig(filename+'.png')
plt.show();
return
def plot_prediction(self,
sample_batch,
prediction,
idx=0,
trans=0.5,
figure=None):
if figure is None:
figure = plt.figure()
figure.tight_layout()
batch_size = len(sample_batch['load_dict'])
assert (idx < batch_size)
# figure.set_size_inches(16, 32)
load_dict = make_tuple(sample_batch['load_dict'][idx])
label = sample_batch['label'][idx].numpy()
image = sample_batch['image'][idx].numpy().transpose(1, 2, 0)
image = 255 * image
image_orig = image.astype(np.uint8)
if self.label_type == 'dense':
image = scp.misc.imresize(image, size=label.shape[:2])
elif self.label_type == 'spatial_2d':
image = scp.misc.imresize(image, size=label.shape[1:])
mask = self.label_coder.getmask(label)
pred = prediction[idx]
# logging.info(pred)
idx = load_dict['idx']
coloured_label = self.label2color(label=label, mask=mask)
coloured_label = trans * image + (1 - trans) * coloured_label
diff_colour = self.coloured_diff(label, pred, mask)
diff_colour = 0.6 * image + 0.4 * diff_colour
coloured_hard = self.pred2color_hard(pred=pred, mask=mask)
coloured_hard = trans * image + (1 - trans) * coloured_hard
ax = figure.add_subplot(2, 2, 1)
ax.set_title('Image #{}'.format(idx))
ax.axis('off')
ax.imshow(image_orig)
ax = figure.add_subplot(2, 2, 2)
ax.set_title('Label')
ax.axis('off')
ax.imshow(coloured_label.astype(np.uint8))
ax = figure.add_subplot(2, 2, 3)
ax.set_title('Failure Map')
ax.axis('off')
ax.imshow(diff_colour.astype(np.uint8))
ax = figure.add_subplot(2, 2, 4)
ax.set_title('Prediction')
ax.axis('off')
ax.imshow(coloured_hard.astype(np.uint8))
return figure
def _parse_list(self, v):
for idx, vv in enumerate(v):
if isinstance(vv, str) and vv.startswith('('):
v[idx] = make_tuple(vv)
return v
def create_tuple(strRGB):
strRGB = strRGB.replace('rgb', '')
arrRGB = make_tuple(strRGB)
return arrRGB
def get_config(seed, shot):
"""
For a given seed and shot, generate a config file based on a template
config file that is used for training/evaluation.
You can extend/modify this function to fit your use-case.
"""
if args.coco:
# COCO
assert args.two_stage, 'Only supports novel weights for COCO now'
if args.novel_finetune:
# Fine-tune novel classifier
ITERS = {
1: (10000, 500),
2: (10000, 1500),
3: (10000, 1500),
5: (10000, 1500),
10: (10000, 2000),
30: (10000, 6000),
}
mode = 'novel'
assert not args.fc and not args.unfreeze
else:
# Fine-tune entire classifier
ITERS = {
1: (14400, 16000),
2: (28800, 32000),
3: (43200, 48000),
5: (72000, 80000),
10: (144000, 160000),
30: (216000, 240000),
}
mode = 'all'
split = temp_split = ''
temp_mode = mode
config_dir = 'configs/COCO-detection'
ckpt_dir = 'checkpoints/coco/faster_rcnn'
base_cfg = '../../Base-RCNN-FPN.yaml'
else:
# PASCAL VOC
assert not args.two_stage, 'Only supports random weights for PASCAL now'
ITERS = {
1: (3500, 4000),
2: (7000, 8000),
3: (10500, 12000),
5: (17500, 20000),
10: (35000, 40000),
}
split = 'split{}'.format(args.split)
mode = 'all{}'.format(args.split)
temp_split = 'split1'
temp_mode = 'all1'
config_dir = 'configs/PascalVOC-detection'
ckpt_dir = 'checkpoints/voc/faster_rcnn'
base_cfg = '../../../Base-RCNN-FPN.yaml'
seed_str = 'seed{}'.format(seed) if seed != 0 else ''
fc = '_fc' if args.fc else ''
unfreeze = '_unfreeze' if args.unfreeze else ''
# Read an example config file for the config parameters
temp = os.path.join(
temp_split,
'faster_rcnn_R_101_FPN_ft{}_{}_1shot{}'.format(fc, temp_mode,
unfreeze))
config = os.path.join(args.root, config_dir, temp + '.yaml')
prefix = 'faster_rcnn_R_101_FPN_ft{}_{}_{}shot{}{}'.format(
fc, mode, shot, unfreeze, args.suffix)
output_dir = os.path.join(args.root, ckpt_dir, seed_str)
os.makedirs(output_dir, exist_ok=True)
save_dir = os.path.join(
args.root,
config_dir,
split,
seed_str,
)
os.makedirs(save_dir, exist_ok=True)
save_file = os.path.join(save_dir, prefix + '.yaml')
configs = load_yaml_file(config)
configs['_BASE_'] = base_cfg
configs['DATASETS']['TRAIN'] = make_tuple(configs['DATASETS']['TRAIN'])
configs['DATASETS']['TEST'] = make_tuple(configs['DATASETS']['TEST'])
if args.coco and not args.novel_finetune:
ckpt_path = os.path.join(output_dir, prefix, 'model_reset_combine.pth')
if not os.path.exists(ckpt_path):
src2 = os.path.join(
output_dir,
'faster_rcnn_R_101_FPN_ft_novel_{}shot{}'.format(
shot, args.suffix),
'model_final.pth',
)
if not os.path.exists(src2):
print('Novel weights do not exist. Please run with the ' + \
'--novel-finetune flag first.')
assert False
combine_cmd = 'python tools/ckpt_surgery.py --coco --method ' + \
'combine --src1 checkpoints/coco/faster_rcnn/faster_rcnn' + \
'_R_101_FPN_base/model_final.pth --src2 {}'.format(src2) + \
' --save-dir {}'.format(os.path.join(output_dir, prefix))
run_cmd(combine_cmd)
assert os.path.exists(ckpt_path)
configs['MODEL']['WEIGHTS'] = ckpt_path
elif not args.coco:
configs['MODEL']['WEIGHTS'] = configs['MODEL']['WEIGHTS'].replace(
'base1', 'base' + str(args.split))
for dset in ['TRAIN', 'TEST']:
configs['DATASETS'][dset] = (configs['DATASETS'][dset][0].replace(
temp_mode, 'all' + str(args.split)), )
configs['DATASETS']['TRAIN'] = (
configs['DATASETS']['TRAIN'][0].replace('1shot',
str(shot) + 'shot') +
('_{}'.format(seed_str) if seed_str != '' else ''), )
configs['SOLVER']['BASE_LR'] = args.lr
configs['SOLVER']['MAX_ITER'] = ITERS[shot][1]
configs['SOLVER']['STEPS'] = (ITERS[shot][0], )
configs['SOLVER']['CHECKPOINT_PERIOD'] = ITERS[shot][1] // args.ckpt_freq
configs['OUTPUT_DIR'] = os.path.join(output_dir, prefix)
if seed != 0:
with open(save_file, 'w') as fp:
yaml.dump(configs, fp)
return save_file, configs
# Remove comments
line = line.split("#")[0].strip()
# Skip blank or comment-only lines
if (not (line)): continue
# If a key is designated as a modifer add it to the appropriate list
p = re.compile('^(S?MOD)\[(\w+)\] (\([^\)]+\))')
m = p.search(line)
if m:
# Extract the groups
modType, modName, keyPos = m.groups()
# Turn keyPos into a tuple
keyPos = make_tuple(keyPos)
# Sort into MOD and SMOD
if modType == 'MOD':
modKeys[modName] = keyPos
elif modType == 'SMOD':
sModKeys[modName] = keyPos
else:
raise Exception("Bad keymap, unknown modifier type")
else:
# Otherwise parse the line
# Separate the key input from the key output and strip of whitespace
lineparts = [part.strip() for part in line.split("=")]
keyOutput = [part.strip() for part in lineparts[1].split("+")]
def load_tuple(val):
return make_tuple(xml_node.attrib[val])
