对象检测标签映射项'str'对象没有属性'item'

Question

我使用Tensorflow Object Detection API训练了一个自定义图像，并使用训练的数据运行了对象检测教程。我有一个与加载标签地图相关的错误。我已经检查了标签图像文件，它似乎与字典内容一致。我不太明白为什么会发生错误。

代码：

# What model to download.
MODEL_NAME = 'new_graph.pb'
# Path to frozen detection graph. This is the actual model that is used for the object detection.
PATH_TO_FROZEN_GRAPH = MODEL_NAME + '/frozen_inference_graph.pb'
# List of the strings that is used to add correct label for each box.
PATH_TO_LABELS = 'training/labelmap.pbtxt'
NUM_CLASSES=3
detection_graph = tf.Graph()
with detection_graph.as_default():
od_graph_def = tf.GraphDef()  
with tf.gfile.GFile(PATH_TO_FROZEN_GRAPH, 'rb') as fid:
serialized_graph = fid.read()
od_graph_def.ParseFromString(serialized_graph)
tf.import_graph_def(od_graph_def, name='')


category_index = label_map_util.convert_label_map_to_categories(PATH_TO_LABELS , max_num_classes=NUM_CLASSES, use_display_name=True) 

错误：

AttributeError                            Traceback (most recent call last)
<ipython-input-27-7acf82e14013> in <module>
1 #category_index = label_map_util.create_category_index_from_labelmap(PATH_TO_LABELS, use_display_name=True)
2 
----> 3 category_index = label_map_util.convert_label_map_to_categories(PATH_TO_LABELS , max_num_classes=NUM_CLASSES)
4 
D:me1eyeNew folder29082020modelsresearchobject_detectionutilslabel_map_util.py in convert_label_map_to_categories(label_map, max_num_classes, use_display_name)
118       })
119     return categories
--> 120   for item in label_map.item:
121     if not 0 < item.id <= max_num_classes:
122       logging.info(
AttributeError: 'str' object has no attribute 'item'

labelmap.pbtxt文件：

item {
id: 1
name: 'Cat'
}
item {
id: 2
name: 'Grabes'
}
item {
id: 3
name: 'Olive'
}

Answer 1

需要更改以下内容：

从utils导入label_map_util

---->从object_detection.utils导入label_map_util

从utils将visualization_utils导入为vis_util

---->从object_detection.utils将visualization_utils导入为vis_util

Answer 2

使用convert_label_map_to_categories时，需要先用load_labelmap加载地图数据。您的代码正在处理文件名，而不是文件数据。

试试这个代码：

label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
category_index = label_map_util.convert_label_map_to_categories(label_map, max_num_classes=NUM_CLASSES, use_display_name=True)