def calculate_image_similarity(self):
print(self.location_id)
print(self.model)
print(self.image_id)
query_vector_index = int(
self.cache.get(self.location_id + '_' + self.model + "_" +
self.image_id))
query_vector = self.reduced_data[self.location_id][query_vector_index]
similar_vectors = []
for locations, images in self.reduced_data.items():
image_index = 0
for image in images:
s = {}
s['query_id'] = self.image_id
s['compare_id'] = self.cache.get(self.location_id + '_' +
self.model + "_" +
str(image_index))
s['value'] = self.math.manhattan_distance(query_vector, image)
similar_vectors = helpers.sort(similar_vectors, s, self.limit,
'value', 1)
image_index = image_index + 1
return similar_vectors
def location_similarity(self):
query_matrix = self.reduced_data[self.location_id]
similar_vectors = []
for k, v in self.reduced_data.items():
s = {}
s['query_id'] = self.location_id
s['compare_id'] = k
s['value'] = self.min_pair_similarity(query_matrix, v)
similar_vectors = helpers.sort(similar_vectors, s, self.limit,
'value', 1)
return similar_vectors
def self_similarity(self, query_id, limit, vector_space):
ids = self.cache.hgetall("dm_"+vector_space)
indices = self.cache.hgetall("idm_"+vector_space)
if vector_space == "location":
query_id = self.cache.hgetall("inverse_location_map")[query_id.encode('utf-8')].decode('utf-8')
print(query_id)
query = self.data[vector_space][int(ids[query_id.encode('utf-8')].decode('utf-8'))]
similar_vectors = []
for i in range(len(self.data[vector_space])):
similarity = {}
similarity['value'] = self.math.cosine_similarity(self.data[vector_space][i], query)
similarity['index'] = str(i)
similarity['id'] = indices[str(i).encode('utf-8')]
similar_vectors = helpers.sort(similar_vectors, similarity, limit, 'value', 0)
return similar_vectors
def find_closest(self, key, v1, v2, k):
minDist = []
average = np.zeros((len(v1[0],)))
for i in range(0, len(v1)):
average = average + v1[i]
average = average / len(v1)
for j in range(0, len(v2)):
val = self.math.cosine_similarity(average, np.ravel(v2[j]))
minDist.append({"distance": val, "term1":i, "term2":j})
minDist = helpers.sort(minDist, {"distance": val, "term1":i, "term2":j}, k, 'distance', 0)
count = 0
l = []
while (count < k):
if minDist[count]["term2"] not in l:
temp = {}
temp['distance'] = minDist[count]['distance']
temp['compare_id'] = self.cache.hgetall("idm_"+key)[str(minDist[count]['term2']).encode('utf-8')]
l.append(temp)
count = count + 1
return l