Deep-Learning-Approach-for-Surface-Defect-Detection
Deep-Learning-Approach-for-Surface-Defect-Detection copied to clipboard
ShuaiLYU
how to save pbfile
I find the function of save pb file is incompleted , I really can't make it ,wish your help , please...
sorry ,I am busy those days. It's easy and You just need to modify the “name” of input and output nodes. I believe you can do it.
yes，I have done that.it works perfectly.thank you.
---Original--- From: "Wslsdx"[email protected] Date: Tue, Jun 4, 2019 13:55 PM To: "Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection"Deep-Learning-Approach-for-Surface-Defect-Detection@noreply.github.com; Cc: "ZOUYIyi"[email protected];"Author"[email protected]; Subject: Re: [Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection] how to save pbfile (#3)
sorry ,I am busy those days. It's easy and You just need to modify the “name” of input and output nodes. I believe you can do it.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.
主要是确定输入输出节点的名字，然后按照网上的例子先做个简单的网络freeze，就可以了。
------------------ 原始邮件 ------------------ 发件人: "gulingfengze"[email protected]; 发送时间: 2019年6月11日(星期二) 上午9:19 收件人: "Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection"Deep-Learning-Approach-for-Surface-Defect-Detection@noreply.github.com; 抄送: "邹易"[email protected];"Mention"[email protected]; 主题: Re: [Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection] howto save pbfile (#3)
@ZOUYIyi Can you tell me how to implement model freeze? Thanks a lot.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
input_name = "Image" output_name = "segment/Sigmoid"
------------------ 原始邮件 ------------------ 发件人: "gulingfengze"[email protected]; 发送时间: 2019年6月11日(星期二) 下午5:13 收件人: "Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection"Deep-Learning-Approach-for-Surface-Defect-Detection@noreply.github.com; 抄送: "邹易"[email protected]; "Mention"[email protected]; 主题: Re: [Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection] howto save pbfile (#3)
@ZOUYIyi What are the input and output node names? I'm not sure.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
@ZOUYIyi Thank you very much for your reply, but I still encountered some problems. I use the following script to go back to the pb file, and according to the output node name you provided, execute the script reminder: AssertionError: segment/Sigmoid is not in graph, the node name is not seen in the code, do I need to add it manually? If so, where do you add it? `import os import argparse import tensorflow as tf from tensorflow.python.framework import graph_util
def freeze_graph(model_dir, output_node_names):
if not tf.gfile.Exists(model_dir):
raise AssertionError(
"Export directory doesn't exists. Please specify an export "
"directory: %s" % model_dir)
if not output_node_names:
print("You need to supply the name of a node to --output_node_names.")
return -1
checkpoint = tf.train.get_checkpoint_state(model_dir)
input_checkpoint = checkpoint.model_checkpoint_path
absolute_model_dir = "/".join(input_checkpoint.split('/')[:-1])
output_graph = absolute_model_dir + "/frozen_model.pb"
clear_devices = True
with tf.Session(graph=tf.Graph()) as sess:
saver = tf.train.import_meta_graph(input_checkpoint + '.meta', clear_devices=clear_devices)
saver.restore(sess, input_checkpoint)
output_graph_def = tf.graph_util.convert_variables_to_constants(
sess,
tf.get_default_graph().as_graph_def(),
output_node_names.split(",")
)
with tf.gfile.GFile(output_graph, "wb") as f:
f.write(output_graph_def.SerializeToString())
print("%d ops in the final graph." % len(output_graph_def.node))
if name == 'main': parser = argparse.ArgumentParser() parser.add_argument("--model_dir", type=str, default="", help="Model folder to export") parser.add_argument("--output_node_names", type=str, default="", help="The name of the output nodes, comma separated.") args = parser.parse_args() freeze_graph(args.model_dir, args.output_node_names)`
here is my script for saving pb file:
////
def save_PbModel(self):
input_name = "Image"
output_name = "segment/Sigmoid"
output_node_names = [input_name,output_name]
print("模型保存为pb格式，输入节点name：{}输出节点name: {}".format(input_name,output_name))
output_graph_def = tf.graph_util.convert_variables_to_constants(
self.__session,
self.__session.graph_def,
output_node_names,
)
with tf.gfile.GFile("zy.pb", "wb") as f:
f.write(output_graph_def.SerializeToString())
print("=> ops written to it")
/////
------------------ 原始邮件 ------------------ 发件人: "gulingfengze"[email protected]; 发送时间: 2019年6月11日(星期二) 晚上6:45 收件人: "Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection"Deep-Learning-Approach-for-Surface-Defect-Detection@noreply.github.com; 抄送: "邹易"[email protected]; "Mention"[email protected]; 主题: Re: [Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection] howto save pbfile (#3)
@ZOUYIyi Thank you very much for your reply, but I still encountered some problems. I use the following script to go back to the pb file, and according to the output node name you provided, execute the script reminder: AssertionError: segment/Sigmoid is not in graph, the node name is not seen in the code, do I need to add it manually? If so, where do you add it? `import os import argparse import tensorflow as tf from tensorflow.python.framework import graph_util
def freeze_graph(model_dir, output_node_names):
"""Extract the sub graph defined by the output nodes and convert
all its variables into constant
Args:
model_dir: the root folder containing the checkpoint state file
output_node_names: a string, containing all the output node's names,
comma separated
"""
if not tf.gfile.Exists(model_dir):
raise AssertionError(
"Export directory doesn't exists. Please specify an export "
"directory: %s" % model_dir)
if not output_node_names: print("You need to supply the name of a node to --output_node_names.") return -1 checkpoint = tf.train.get_checkpoint_state(model_dir) input_checkpoint = checkpoint.model_checkpoint_path absolute_model_dir = "/".join(input_checkpoint.split('/')[:-1]) output_graph = absolute_model_dir + "/frozen_model.pb" clear_devices = True with tf.Session(graph=tf.Graph()) as sess: saver = tf.train.import_meta_graph(input_checkpoint + '.meta', clear_devices=clear_devices) saver.restore(sess, input_checkpoint) output_graph_def = tf.graph_util.convert_variables_to_constants( sess, tf.get_default_graph().as_graph_def(), output_node_names.split(",") ) with tf.gfile.GFile(output_graph, "wb") as f: f.write(output_graph_def.SerializeToString()) print("%d ops in the final graph." % len(output_graph_def.node))
if name == 'main':
parser = argparse.ArgumentParser()
parser.add_argument("--model_dir", type=str, default="", help="Model folder to export")
parser.add_argument("--output_node_names", type=str, default="", help="The name of the output nodes, comma separated.")
args = parser.parse_args()
freeze_graph(args.model_dir, args.output_node_names)`
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
@ZOUYIyi Thank you very much for your help. I modified my model freezing script, successfully exported pb file, and tested well on my test script.
@gulingfengze Could you please send me the freezing script and test script?I am new here. and I want to build an interface , the input is an image and out put is the predicted label
@HuitMahoon I'm happy to do it all， freeze script test scrip
@gulingfengze @HuitMahoon 请问你们在自己数据上面跑过吗，能否分享你们对数据标注和结果预测的经验？感谢
input_name = "Image" output_name = "segment/Sigmoid" … ------------------ 原始邮件 ------------------ 发件人: "gulingfengze"[email protected]; 发送时间: 2019年6月11日(星期二) 下午5:13 收件人: "Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection"Deep-Learning-Approach-for-Surface-Defect-Detection@noreply.github.com; 抄送: "邹易"[email protected]; "Mention"[email protected]; 主题: Re: [Wslsdx/Deep-Learning-Approach-for-Surface-Defect-Detection] howto save pbfile (#3) @ZOUYIyi What are the input and output node names? I'm not sure. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
input_name = 'Image', output_name = "segment/Sigmoid" , the globalvariables are freezed, but when the pb files is loaded with opencv, readNetFromTensorflow， error happens( something like "can not detetmine the number of batchnorm layers")
