使用Inception v3实现图像识别

约定不等于承诺〃 2023-07-23 10:59 42阅读 0赞

官方相关模型及文件下载：[http://download.tensorflow.org/models/image/imagenet/inception-2015-12-05.tgz][http_download.tensorflow.org_models_image_imagenet_inception-2015-12-05.tgz]  
下载的文件列表：  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2F5YW5nYW5uOTE1_size_16_color_FFFFFF_t_70]  
里面有训练好的pb模型，以及标签特征对应文件  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2F5YW5nYW5uOTE1_size_16_color_FFFFFF_t_70 1]

载入相关库：

import tensorflow as tf
    import os
    import numpy as np
    import re
    from PIL import Image
    import matplotlib.pyplot as plt

code：

class NodeLookup(object):
        def __init__(self):
        #加载对应的分类文件
            label_path = "inception_model/imagenet_2012_challenge_label_map_proto.pbtxt"
            uid_path = "inception_model/imagenet_synset_to_human_label_map.txt"
            self.node_lookup = self.load(label_path, uid_path)
        
        def load(self, label_path, uid_path):
        #读取pbtxt文件，按行读取
            proto_label_line = tf.gfile.GFile(label_path).readlines()
            #将读取的内容以字典的方式存储，key-value
            node_to_label_uid = { }
            #遍历每一行
            for line in proto_label_line:
            #去掉换行符
                line = line.strip('\n')
                #如果每一行以target_class开头：则对应的内容为 target_class: 442
                if line.startswith(" target_class:"):
                    target_class = int(line.split(": ")[1])#442
                if line.startswith(" target_class_string:"):
                #包含引号
                    target_class_string = line.split(": ")[1]
                    # target_class_string: "n01494475"，取中间的字母数字部分，不包含引号
                    node_to_label_uid[target_class] = target_class_string[1:-1]
            
            proto_uid_line = tf.gfile.GFile(uid_path).readlines()
            uid_to_label_human = { }
            for line in proto_uid_line:
                line = line.strip('\n')
                #分割
                parse_items = line.split("\t")
                uid = parse_items[0]
                target = parse_items[1]
                uid_to_label_human[uid] = target
                #uid_to_label_human[parse_items[0]] = parse_items[1]
                
            #重新对应如种类编号442对应类别cat（假设）
            node_id_to_name = { }
            for key, val in node_to_label_uid.items():
                name = uid_to_label_human[val]
                node_id_to_name[key] = name
            return node_id_to_name
        
        def id_to_string(self, node_id):
            if node_id not in self.node_lookup:
                return ''
            return self.node_lookup[node_id]
    
    #读取模型，重新创建相应的图，固定书写模式
    with tf.gfile.GFile("inception_model/classify_image_graph_def.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:
        softmax_tensor = sess.graph.get_tensor_by_name("softmax:0")
        for root, dirs, files in os.walk("images/"):
            for file in files:
                image_data  = tf.gfile.GFile(os.path.join(root, file), 'rb').read()
                predict = sess.run(softmax_tensor, { "DecodeJpeg/contents:0":image_data})#图片是jpg格式
                predict = np.squeeze(predict)#将结果转为一维数据
                print(predict.shape)
                
                image_path = os.path.join(root, file)
                print(image_path)
                img = Image.open(image_path)
                plt.imshow(img)
                plt.axis("off")
                plt.show()
                
                top_pre = predict.argsort()[-3:][::-1]#升序排列，选择最后三个数据，再倒叙，即降序
                node_lookup = NodeLookup()
                for node_id in top_pre:
                    human_string = node_lookup.id_to_string(node_id)
                    score = predict[node_id]
                    print("%s (score=%.5f)" %(human_string, score))
                print("*************************")

结果：  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2F5YW5nYW5uOTE1_size_16_color_FFFFFF_t_70 2]

#### bug ####

1、readline()&readlines(),这里使用readlines(),否则只读取一行数据

---------------------------------------------------------------------------
    IndexError                                Traceback (most recent call last)
    <ipython-input-21-74081abf60d9> in <module>
         74             #对预测结果排序，从小到大排序，取最后几个，[::-1]：对最后的五个概率值取倒叙
         75             top_k = predictions.argsort()[-5:][: :-1]
    ---> 76             node_lookup = NodeLookUp()
         77             for node_id in top_k:
         78                 #获取分类名称
    
    <ipython-input-21-74081abf60d9> in __init__(self)
          4         label_lookup_path = "inception_model/imagenet_2012_challenge_label_map_proto.pbtxt"
          5         uid_lookup_path = "inception_model/imagenet_synset_to_human_label_map.txt"
    ----> 6         self.node_lookup = self.load(label_lookup_path, uid_lookup_path)
          7 
          8     #载入数据进行处理
    
    <ipython-input-21-74081abf60d9> in load(self, label_lookup_path, uid_lookup_path)
         19             #用键值对存储数据
         20             uid = parsed_items[0]
    ---> 21             human_string = parsed_items[1]
         22             #保存id和字符串为对应的映射关系
         23             uid_to_human[uid] = human_string
    
    IndexError: list index out of range

2、载入模型时没有对其采用二进制读取，‘rb’

---------------------------------------------------------------------------
    UnicodeDecodeError                        Traceback (most recent call last)
    <ipython-input-27-052cb1c2f17f> in <module>
         40 with tf.gfile.GFile("inception_model/classify_image_graph_def.pb") as f:
         41     graph_def = tf.GraphDef()
    ---> 42     graph_def.ParseFromString(f.read())
         43     tf.import_graph_def(graph_def, name="")
        UnicodeDecodeError: 'utf-8' codec can't decode byte 0xbb in position 1: invalid start byte

‘r’:默认值，表示从文件读取数据。  
‘w’:表示要向文件写入数据，并截断以前的内容  
‘a’:表示要向文件写入数据，添加到当前内容尾部  
‘r+’:表示对文件进行可读写操作（删除以前的所有数据）  
‘r+a’：表示对文件可进行读写操作（添加到当前文件尾部）  
‘b’:表示要读写二进制数据

3、未定义，tab，代码格式问题

if line.startswith(" target_class_string:"):
                    target_class_string = line.split(": ")[1]
    node_to_label_uid[target_class] = target_class_string[1:-1]
    
    <ipython-input-28-ecfdf82b86d1> in load(self, label_path, uid_path)
         14             if line.startswith(" target_class_string:"):
         15                 target_class_string = line.split(": ")[1]
    ---> 16             node_to_label_uid[target_class] = target_class_string
         17 
         18         proto_uid_line = tf.gfile.GFile(uid_path).readlines()
    
    UnboundLocalError: local variable 'target_class_string' referenced before assignment

[http_download.tensorflow.org_models_image_imagenet_inception-2015-12-05.tgz]: http://download.tensorflow.org/models/image/imagenet/inception-2015-12-05.tgz
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2F5YW5nYW5uOTE1_size_16_color_FFFFFF_t_70]: /images/20230528/5606abb43fbd4e7d9746c20f5a101067.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2F5YW5nYW5uOTE1_size_16_color_FFFFFF_t_70 1]: /images/20230528/cb2c36241e1b43bda0778eb622d0c707.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2F5YW5nYW5uOTE1_size_16_color_FFFFFF_t_70 2]: /images/20230528/dac11e60260444149aa4277201804be2.png