使用张量流程的错误图形Def不能大于2GB

Question

我使用初始v3模型imageNet形状与张量流进行图像分类。该程序旨在对单个图像进行分类，因此我试图对其进行修改以对测试图像数据库进行分类。它归类图像很好，但在约20幅图像到达返回我下面的错误：File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 1912, in as_graph_def raise ValueError("GraphDef cannot be larger than 2GB.") ValueError: GraphDef cannot be larger than 2GB.

下面是我修改的图像标签代码：

 # -*- coding: utf-8 -*- 

import os, sys 
import time 
import tensorflow as tf 

def chargement_image(path): 
    image = [] 
    image = os.listdir(path) 

    return image 

resultat = [] 
best = [] 
nbr = 0 
som = 0 
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' 

start_time = time.time() 

# Chargement de la base de test 
test_path = sys.argv[1] 

list_img = chargement_image(test_path) 
for i in range(len(list_img)): 
    image_path = test_path+list_img[i] 

# Read in the image_data 
    image_data = tf.gfile.FastGFile(image_path, 'rb').read() 

# Loads label file, strips off carriage return 
    label_lines = [line.rstrip() for line 
        in tf.gfile.GFile("retrained_labels.txt")] 

# Unpersists graph from file 
    with tf.gfile.FastGFile("retrained_graph.pb", 'rb') as f: 
     graph_def = tf.GraphDef() 
     graph_def.ParseFromString(f.read()) 
     tf.import_graph_def(graph_def, name='') 

    with tf.Session() as sess: 
    # Feed the image_data as input to the graph and get first prediction 
     softmax_tensor = sess.graph.get_tensor_by_name('final_result:0') 

     predictions = sess.run(softmax_tensor, \ 
      {'DecodeJpeg/contents:0': image_data}) 
     #print(len(predictions)) 
    # Trier pour afficher les étiquettes de la première prédiction par ordre de bon taux de classement 
     top_k = predictions[0].argsort()[-len(predictions[0]):][::-1] 
     #print(top_k) 

     for node_id in top_k: 
      human_string = label_lines[node_id] 
      score = predictions[0][node_id] 
     resultat.append(score) 
      print('%s (score = %.5f)' % (human_string, score)) 
     #print(score) 

    nbr += 1 
    best.append(resultat[0]) 
    del resultat[:] 
    #print(best) 
    print(nbr) 
    print("=========================================") 
#print(best) 
#print(nbr) 
for i in range(len(best)): 
    som += best[i] 
taux_precision = float(100. * som/nbr) 
print 'Precision: ' + str(taux_precision) + '%' 
print("--- %s seconds ---" % (time.time() - start_time)) 

Answer 1

尽量把下面线外的for循环：

# Loads label file, strips off carriage return 
label_lines = [line.rstrip() for line 
       in tf.gfile.GFile("retrained_labels.txt")] 

# Unpersists graph from file 
with tf.gfile.FastGFile("retrained_graph.pb", 'rb') as f: 
    graph_def = tf.GraphDef() 
    graph_def.ParseFromString(f.read()) 
    tf.import_graph_def(graph_def, name='')