class Rtree2D(object):
"""Wrapper of `rtree.Index` for supporting friendly 2d operations.
Also forces the uniqueness of the `id` parameter, which is different from
the rtree module's behavior.
"""
def __init__(self):
self._index = Rtree()
self._locations = {}
@staticmethod
def to_coords(location):
return (location[0], location[1], location[0], location[1])
def keys(self):
return self._locations.keys()
def get(self, id, objects=False):
return self._locations.get(id)
def set(self, id, location=None, state=None, obj=None):
# Clean up previous value first if any
old = self._locations.get(id)
if old is not None:
self._index.delete(id, self.to_coords(old[0]))
if old and state is None:
state = old[1]
if old and location is None:
location = old[0]
self._locations[id] = (location, state)
self._index.insert(id, self.to_coords(location), obj=obj)
def update(self, id, state, obj=None):
old = self._locations.get(id)
if old is not None:
self._index.delete(id, self.to_coords(old[0]))
self._locations[id] = (old[0], state)
self._index.insert(id, self.to_coords(old[0]), obj=obj)
def remove(self, id):
self._index.delete(id, self.to_coords(self._locations[id][0]))
del self._locations[id]
def nearest(self, location, count=1, objects=False, max_distance=None):
ids = self._index.nearest(self.to_coords(location), num_results=count,
objects=objects)
if max_distance is not None:
ids = [(id_, self._locations[id_][1]) for id_ in ids
if distance(self._locations[id_][0], location) <= max_distance]
return ids
hasData = property( lambda self: RTreePage.objects(name=self.name).first() is not None )
if __name__=='__main__':
settings = Property()
settings.writethrough= True
settings.buffering_capacity=1
storage = MongoStorage('test')
storage.clear()
r = Rtree( storage, properties=settings)
r.add(123,(0,0,1,1))
print "test 1 should be true"
item = list(r.nearest((0,0), 1, objects=True))[0]
print item.id
print r.valid()
print "test 2 should be true"
r.delete(123, (0,0,1,1))
print r.valid()
print "test 3 should be true"
r.clearBuffer()
print r.valid()
del r
print "test 4 should be false"
storage.clear()
lambda self: RTreePage.objects(name=self.name).first() is not None)
if __name__ == '__main__':
settings = Property()
settings.writethrough = True
settings.buffering_capacity = 1
storage = MongoStorage('test')
storage.clear()
r = Rtree(storage, properties=settings)
r.add(123, (0, 0, 1, 1))
print "test 1 should be true"
item = list(r.nearest((0, 0), 1, objects=True))[0]
print item.id
print r.valid()
print "test 2 should be true"
r.delete(123, (0, 0, 1, 1))
print r.valid()
print "test 3 should be true"
r.clearBuffer()
print r.valid()
del r
print "test 4 should be false"
storage.clear()
def searching(feature_vectors_database, feature_vectors_retrieval,
similarity_metric, image_paths, retrieval_number, file,
list_of_parameters, feature_extraction_method, path_database):
'''
feature_vectors: atriutos calculados
labels: label de cada classe
similarity_metric: qual medida utilizar
recuperados as k imagens com menor distancia. Se k = 0, entao o valor eh
setado como sendo o tamanho da classe da imagem
'''
#name to save the pickle file
parameters_name = ""
for parameter in list_of_parameters:
parameters_name = parameters_name + "_" + parameter
file = path_database + "features/sortingRTree" + "_" + feature_extraction_method + parameters_name + '_' + similarity_metric
feature_vectors_retrieval = preprocessing.scale(feature_vectors_retrieval)
if not (os.path.isfile(file + '.dat')):
#normalize signatures
feature_vectors_database = preprocessing.scale(
feature_vectors_database)
# Create a N-Dimensional index
p = index.Property()
p.dimension = feature_vectors_database.shape[1]
idx = index.Index(file, properties=p)
# Create the tree
for i, vector in enumerate(feature_vectors_database):
idx.add(i, vector.tolist())
#save_format = idx.dumps(idx)
#with open(file, 'wb') as handle:
#pickle.dump(save_format, handle)
else:
# Create a N-Dimensional index
p = index.Property()
p.dimension = feature_vectors_database.shape[1]
idx = Rtree(file, properties=p)
#with open(file, 'rb') as handle:
# idx = pickle.load(handle)
# Find closests pair for the first N points
########### debug this part ###########
small_distances = []
for id1, query in enumerate(feature_vectors_retrieval):
nearest = list(idx.nearest(query.tolist(), retrieval_number))
small_distances.append(nearest)
result = []
for cont1, i in enumerate(small_distances):
aux = []
for j in i:
aux.append(image_paths[j])
result.append(aux)
return result
