def second_check1(input):
'''
第二层分类器1
:param input: 输入特征向量(矩阵)/标签
:return: 具体类别, numpy.ndarray
'''
dict1 = {0: 7.174541, 1: 17.936353, 2: 20.0}
pb_file_path = r'/home/xiaosong/桌面/pny相关数据/model/full_model_1'
g2 = tf.Graph()
with g2.as_default():
init = tf.global_variables_initializer()
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options),
graph=g2) as sess:
sess.run(init)
use_pb(sess_new=sess,
pb_file_path=pb_file_path,
file_suffix=r'/second_classifier_1')
x_f = import_ops(sess_new=sess, op_name='placeholder/x_f')
x_l = import_ops(sess_new=sess, op_name='placeholder/x_l')
is_training = import_ops(sess_new=sess,
op_name='placeholder/is_training')
output = import_ops(sess_new=sess, op_name='dnn/Softmax')
r_classify = sess.run(output,
feed_dict={
x_f: input[:, :4],
x_l: input[:, 4:-1],
is_training: False
})
r_classify = np.argmax(a=r_classify, axis=1)
#按照dict1中的映射将类别标记替换为实际半径
r_finally = []
for i in range(r_classify.shape[0]):
r_finally.append(dict1[r_classify[i]])
return r_finally
def first_check(input):
'''
第一层分类器
:param input: 输入特征向量(矩阵)/标签
:return: 半径所属大类别, numpy.ndarray
'''
pb_file_path = r'/home/xiaosong/桌面/pny相关数据/model/full_model'
g1 = tf.Graph()
with g1.as_default():
init = tf.global_variables_initializer()
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options),
graph=g1) as sess:
sess.run(init)
use_pb(sess_new=sess,
pb_file_path=pb_file_path,
file_suffix=r'/first_classifier')
x_f = import_ops(sess_new=sess, op_name='placeholder/x_f')
x_l = import_ops(sess_new=sess, op_name='placeholder/x_l')
is_training = import_ops(sess_new=sess,
op_name='placeholder/is_training')
output = import_ops(sess_new=sess, op_name='dnn/Softmax')
r_classify = sess.run(output,
feed_dict={
x_f: input[:, :4],
x_l: input[:, 4:-1],
is_training: False
})
r_classify = np.argmax(a=r_classify, axis=1)
return r_classify
def second_check1(input):
'''
第二层分类器1
:param input: 输入特征向量(矩阵)/标签
:return: 具体类别, numpy.ndarray
'''
dict1 = {0: 7.17, 1: 17.93, 2: 20}
pb_file_path = r'/home/xiaosong/桌面/pny相关数据/model/full_model_1'
g2 = tf.Graph()
with g2.as_default():
init = tf.global_variables_initializer()
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options),
graph=g2) as sess:
sess.run(init)
use_pb(sess_new=sess,
pb_file_path=pb_file_path,
file_suffix=r'/second_classifier_1')
x_f = import_ops(sess_new=sess, op_name='placeholder/x_f')
x_l = import_ops(sess_new=sess, op_name='placeholder/x_l')
is_training = import_ops(sess_new=sess,
op_name='placeholder/is_training')
output = import_ops(sess_new=sess, op_name='dnn/Softmax')
r_classify = sess.run(output,
feed_dict={
x_f: input[:, :4],
x_l: input[:, 4:],
is_training: False
})
# tf.summary.FileWriter('log/second_graph1', sess.graph)
r_classify = np.argmax(a=r_classify, axis=1)
r_finally = dict1[r_classify]
return r_finally
def second_check2(input):
'''
第二层分类器1
:param input: 输入特征向量(矩阵)/标签
:return: 具体类别, numpy.ndarray
'''
dict2 = {
0: 0.05,
1: 3.261155,
2: 4.484088,
3: 23.915137,
4: 35.872705,
5: 71.745411,
6: 107.618116,
7: 143.490822,
8: 179.363527,
9: 215.236233,
10: 251.108938
}
pb_file_path = r'/home/xiaosong/桌面/pny相关数据/model/full_model_2'
g3 = tf.Graph()
with g3.as_default():
init = tf.global_variables_initializer()
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options),
graph=g3) as sess:
sess.run(init)
use_pb(sess_new=sess,
pb_file_path=pb_file_path,
file_suffix=r'/second_classifier_2')
x_f = import_ops(sess_new=sess, op_name='placeholder/x_f')
x_l = import_ops(sess_new=sess, op_name='placeholder/x_l')
is_training = import_ops(sess_new=sess,
op_name='placeholder/is_training')
output = import_ops(sess_new=sess, op_name='sub_dnn/Softmax')
r_classify = sess.run(output,
feed_dict={
x_f: input[:, :4],
x_l: input[:, 4:-1],
is_training: False
})
r_classify = np.argmax(a=r_classify, axis=1)
# 按照dict2中的映射将类别标记替换为实际半径
r_finally = []
for i in range(r_classify.shape[0]):
r_finally.append(dict2[r_classify[i]])
return r_finally
def second_check2(input):
'''
第二层分类器1
:param input: 输入特征向量(矩阵)/标签
:return: 具体类别, numpy.ndarray
'''
dict2 = {
0: 0.05,
1: 0.672281,
2: 0.924386,
3: 1.479017,
4: 3.697543,
5: 4.930057,
6: 20.0,
7: 22.185257,
8: 36.975428,
9: 44.370514
}
pb_file_path = r'/home/xiaosong/桌面/oldenburg相关数据/model/full_model_2'
g3 = tf.Graph()
with g3.as_default():
init = tf.global_variables_initializer()
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options),
graph=g3) as sess:
sess.run(init)
use_pb(sess_new=sess,
pb_file_path=pb_file_path,
file_suffix=r'/second_classifier_2')
x_f = import_ops(sess_new=sess, op_name='placeholder/x_f')
x_l = import_ops(sess_new=sess, op_name='placeholder/x_l')
is_training = import_ops(sess_new=sess,
op_name='placeholder/is_training')
output = import_ops(sess_new=sess, op_name='sub_dnn/Softmax')
r_classify = sess.run(output,
feed_dict={
x_f: input[:, :4],
x_l: input[:, 4:-1],
is_training: False
})
tf.summary.FileWriter('log/second_graph2', sess.graph)
r_classify = np.argmax(a=r_classify, axis=1)
# 按照dict2中的映射将类别标记替换为实际半径
r_finally = []
for i in range(r_classify.shape[0]):
r_finally.append(dict2[r_classify[i]])
return r_finally
def second_check2(input):
'''
第二层分类器1
:param input: 输入特征向量(矩阵)/标签
:return: 具体类别, numpy.ndarray
'''
dict2 = {
0: 0.05,
1: 3.26,
2: 4.48,
3: 23,
4: 35,
5: 71,
6: 107,
7: 143,
8: 179,
9: 215,
10: 251
}
pb_file_path = r'/home/xiaosong/桌面/pny相关数据/model/full_model_2'
g3 = tf.Graph()
with g3.as_default():
init = tf.global_variables_initializer()
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options),
graph=g3) as sess:
sess.run(init)
use_pb(sess_new=sess,
pb_file_path=pb_file_path,
file_suffix=r'/second_classifier_2')
x_f = import_ops(sess_new=sess, op_name='placeholder/x_f')
x_l = import_ops(sess_new=sess, op_name='placeholder/x_l')
is_training = import_ops(sess_new=sess,
op_name='placeholder/is_training')
output = import_ops(sess_new=sess, op_name='sub_dnn/Softmax')
r_classify = sess.run(output,
feed_dict={
x_f: input[:, :4],
x_l: input[:, 4:],
is_training: False
})
tf.summary.FileWriter('log/second_graph2', sess.graph)
r_classify = np.argmax(a=r_classify, axis=1)
# print(r_classify)
r_finally = dict2[r_classify[0]]
return r_finally