def load_feature( feature_proto_path ): fd = rwOperate.read_dict_des(feature_proto_path ) feature_list = [] for key in fd.keys(): feature_list.append( fd[key] ) feature_train = np.array( feature_list, dtype=np.float32 ) return fd.keys(), feature_train
def load_feature(feature_proto_path):
#从存储文件中提取vlad特征或这cnn特征
fd = rwOperate.read_dict_des(feature_proto_path)
feature_list = []
for key in fd.keys():
feature_list.append(fd[key])
all_feature = np.array(feature_list, dtype=np.float32)
return fd.keys(), all_feature
def demo2():
# fd = rwOperate.read_dict_des( os.path.join('/home/kenneth/gitstore/datafolder/com', 'image_cnn_dict.feature') )
fd = rwOperate.read_dict_des( os.path.join('/home/kenneth/gitstore/datafolder/com', 'descriptors_dict.vlad') )
print( fd.keys() )
for key in fd.keys():
print( fd[key][0] )
print( len(fd[key]) )
print( type( fd[key]))
def demo3():
#test for save_dict_des and read_dict_des function which only save the des without keypoint
#these two fucntion can use to save deeplearn feature or vald feature
dicte = {}
dicte['123'] = [1, 2, 3, 4]
dicte['345'] = [12, 3, 4, 56, 7, 9]
rwOperate.save_dict_des(dicte, './test/feature.cnn_feature')
a = rwOperate.read_dict_des('./test/feature.cnn_feature')
for key in a.keys():
print(key)
print(type(a[key]))
print(a[key])
def cnn_feature_extract_demo():
# image_path = '../datafolder/CompanyStandardLabel'
# image_path = './TestLabelImage'
image_path = '/home/kenneth/gitstore/datafolder/img2000'
net = cnn_feature_extract.load_net()
feature_dict = cnn_feature_extract.extract_feature(net, image_path)
# print( feature_dict )
rwOperate.save_dict_des(
feature_dict,
'/home/kenneth/gitstore/datafolder/img2000_feature/image_cnn_dict.feature'
)
fd = rwOperate.read_dict_des(
'/home/kenneth/gitstore/datafolder/img2000_feature/image_cnn_dict.feature'
)
print(fd)
def load_VLAD_from_proto(descriptor_dict_path):
if not os.path.exists(descriptor_dict_path):
print('descriptor_dict_path is not exist!')
descriptors_dict = rwOperate.read_dict_des(descriptor_dict_path)
return descriptors_dict