【AI小疑问】如何加载PB文件并推理预测

开头name

 

 

 

 

结尾name

 

 

 

 

pb加载方式

import tensorflow as tf
from tensorflow.python.platform import gfile
import cv2
import numpy
from keras.applications.xception import preprocess_input

pb_file_path = '/Users/jack/Documents/DaiCode/T4/trans_model'

config = tf.ConfigProto()
sess = tf.Session(config=config)
with gfile.FastGFile('/Users/jack/Documents/DaiCode/T4/trans_model/false_positive3.pb', 'rb') as f:
     graph_def = tf.GraphDef()
     graph_def.ParseFromString(f.read())
     sess.graph.as_default()
     tf.import_graph_def(graph_def, name='')
 
 # 获取输入tensor
x = tf.get_default_graph().get_tensor_by_name("input_1:0")
print("input:", x)
# 获取预测tensor
pred = tf.get_default_graph().get_tensor_by_name("output_1:0")  # mobilenet_v2
print(pred)

打印的输入和输出 

input: Tensor("input_1:0", shape=(?, 240, 240, 3), dtype=float32)
Tensor("output_1:0", shape=(8,), dtype=float32)

 

240*240 3通道 的图片向量 

一位 8 的输出

 

h5加载方式

from keras.models import load_model

# 加载模型的位置
model = load_model('/Users/jack/Documents/DaiCode/T4/false_positive3.h5',
                   custom_objects={'optimizer': optimizer})

 

h5 推理

prediction = model.predict(plate_img)

   pb 推理

prediction = sess.run(pred, feed_dict={x: _plate_img})
[0.98,0.1,0.1,0.2,0.3,0.4,0.6,0.1]