def callback(recognizer, audio):
try:
sents = recognizer.recognize_google(audio , language = "en")
print(sents)
sents = precompute(sents)
print(sents)
for relation in listeRelation:
rels = relation.extract(sents)
for rel in rels:
rel.post();
except sr.UnknownValueError:
print("Google Speech Recognition could not understand audio")
except sr.RequestError:
print("Could not request results from Google Speech Recognition service")
import speech_recognition as sr
import nltk
from server import post_request
from precompute import precompute
from supportedRelations import *
from supportedRelations import listeRelation, dic
# sents = "The capital of France is Paris."
# sents = "France's capital is Paris."
# sents = "France's capital is Palaiseau."
# sents = "Paris is the capital of France."
# sents = "There are 70 million people in France."
# sents = "There are more than 10 million people in France."
# sents = "There are less than 10 million people in France."
# sents = "population density of France is 100 inhabitants per square kilometer"
# sents = "The gdp in France is 10 billion dollars"
# sents = "France's import is 10 billion dollars"
sents = "France has population density greater than 100 people per kilometer square"
print("Analysed sentence : ", sents)
sents = precompute(sents)
print("Precomputed sentence : ", sents)
for relation in listeRelation:
rels = relation.extract(sents)
for rel in rels:
rel.post();
params_file_name = "{}_params_{}_{}_{}.npy".format(file_name, NUM_FOV, NUM_H, NUM_V)
t1 = time.time()
if os.path.exists(params_file_name):
params = np.load(params_file_name)
else:
params = get_params(MIN_FOV, MAX_FOV, NUM_FOV, NUM_H, NUM_V)
np.save(params_file_name, params)
t2 = time.time()
print(params.shape)
print("Got params")
table_file_name = "{}_table_{}_{}_{}_{}_{}.npy".format(file_name, table_type, SAMPLE_RATE, NUM_FOV, NUM_H, NUM_V)
if os.path.exists(table_file_name):
heat_table = np.load(table_file_name)
elif table_type == 0:
heat_table = precompute(data, file_name, SAMPLE_RATE, params, write_masks=True)
np.save(file_name, heat_table)
elif table_type == 1:
temp_name = "{}_table_{}_{}_{}_{}_{}.npy".format(file_name, REGULAR, SAMPLE_RATE, NUM_FOV, NUM_H, NUM_V)
if os.path.exists(temp_name):
print("triggered")
heat_table = np.load(temp_name)
else:
heat_table = precompute(data, file_name, SAMPLE_RATE, params, write_masks=True)
np.save(temp_name, heat_table)
heat_table = normalize_table(heat_table)
np.save(table_file_name, heat_table)
t3 = time.time()
print("Parameter time: {}".format(t2 - t1))
if process_type == "OPF+OPF" or process_type == "OPF+SVM":
distance_type = str(data[7])
distance_param = str(data[8])
print "Num K: ", num_k
print "Thumbnail size: ", thumbnail_size
print "Feature type: ", feature_type
print "Descriptor type: ", descriptor_type
print "Precomputing descriptors"
pool = None
if feature_type == "OVERFEAT" or descriptor_type == "OVERFEAT":
pool = multiprocessing.Pool(processes=1)
else:
pool = multiprocessing.Pool()
precompute.precompute(train_path, test_path, feature_type, descriptor_type, thumbnail_size, pool)
pool.close()
pool.terminate()
if process_type == "KMeans+SVM":
process_kmeans_svm(num_k=num_k, n_sample_images=n_sample_images,
n_sample_descriptors=n_sample_descriptors, thumbnail_size=thumbnail_size,
feature_type=feature_type, descriptor_type=descriptor_type)
elif process_type == "KMeans+OPF":
process_kmeans_opf(num_k=num_k, n_sample_images=n_sample_images,
n_sample_descriptors=n_sample_descriptors, thumbnail_size=thumbnail_size,
feature_type=feature_type, descriptor_type=descriptor_type)
elif process_type == "OPF+OPF":
process_opf_opf(num_k=num_k, n_sample_images=n_sample_images,
n_sample_descriptors=n_sample_descriptors, thumbnail_size=thumbnail_size,
feature_type=feature_type, descriptor_type=descriptor_type,
if process_type == "OPF+OPF" or process_type == "OPF+SVM":
distance_type = str(data[7])
distance_param = str(data[8])
print "Num K: ", num_k
print "Thumbnail size: ", thumbnail_size
print "Feature type: ", feature_type
print "Descriptor type: ", descriptor_type
print "Precomputing descriptors"
pool = None
if feature_type == "OVERFEAT" or descriptor_type == "OVERFEAT":
pool = multiprocessing.Pool(processes=1)
else:
pool = multiprocessing.Pool()
precompute.precompute(train_path, test_path, feature_type, descriptor_type,
thumbnail_size, pool)
pool.close()
pool.terminate()
if process_type == "KMeans+SVM":
process_kmeans_svm(num_k=num_k,
n_sample_images=n_sample_images,
n_sample_descriptors=n_sample_descriptors,
thumbnail_size=thumbnail_size,
feature_type=feature_type,
descriptor_type=descriptor_type)
elif process_type == "KMeans+OPF":
process_kmeans_opf(num_k=num_k,
n_sample_images=n_sample_images,
n_sample_descriptors=n_sample_descriptors,
thumbnail_size=thumbnail_size,
