Question

我尝试在Tensorflow 2上运行Tensorflow对象检测API，但出现此错误，有人可以找到解决方案吗？

代码：

加载程序

def load_model(model_name):
  base_url = 'http://download.tensorflow.org/models/object_detection/'
  model_file = model_name + '.tar.gz'
  model_dir = tf.keras.utils.get_file(
    fname=model_name, 
    origin=base_url + model_file,
    untar=True)

  model_dir = pathlib.Path(model_dir)/"saved_model"

  model = tf.saved_model.load(str(model_dir))
  model = model.signatures['serving_default']

  return model

正在加载标签图

标签将地图索引映射到类别名称，因此当我们的卷积网络预测为5时，我们知道这对应于飞机。这里我们使用内部实用程序函数，但是任何返回将整数映射到适当的字符串标签的字典的方法都可以

# List of the strings that is used to add correct label for each box.
PATH_TO_LABELS = 'data/mscoco_label_map.pbtxt'
category_index = label_map_util.create_category_index_from_labelmap(PATH_TO_LABELS, use_display_name=True)

为简单起见，我们将对2张图像进行测试：

# If you want to test the code with your images, just add path to the images to the TEST_IMAGE_PATHS.
PATH_TO_TEST_IMAGES_DIR = pathlib.Path('test_images')
TEST_IMAGE_PATHS = sorted(list(PATH_TO_TEST_IMAGES_DIR.glob("*.jpg")))
TEST_IMAGE_PATHS

检测

加载对象检测模型：

model_name = 'ssd_mobilenet_v1_coco_11_06_2017'
detection_model = load_model(model_name)

我得到了这个错误

OSError                                   Traceback (most recent call last)
<ipython-input-7-e89d9e690495> in <module>
      1 model_name = 'ssd_mobilenet_v1_coco_11_06_2017'
----> 2 detection_model = load_model(model_name)

<ipython-input-4-f8a3c92a04a4> in load_model(model_name)
      9   model_dir = pathlib.Path(model_dir)/"saved_model"
     10 
---> 11   model = tf.saved_model.load(str(model_dir))
     12   model = model.signatures['serving_default']
     13 

D:\Anaconda\lib\site-packages\tensorflow_core\python\saved_model\load.py in load(export_dir, tags)
    515     ValueError: If `tags` don't match a MetaGraph in the SavedModel.
    516   """
--> 517   return load_internal(export_dir, tags)
    518 
    519 

D:\Anaconda\lib\site-packages\tensorflow_core\python\saved_model\load.py in load_internal(export_dir, tags, loader_cls)
    524     # sequences for nest.flatten, so we put those through as-is.
    525     tags = nest.flatten(tags)
--> 526   saved_model_proto = loader_impl.parse_saved_model(export_dir)
    527   if (len(saved_model_proto.meta_graphs) == 1
    528       and saved_model_proto.meta_graphs[0].HasField("object_graph_def")):

D:\Anaconda\lib\site-packages\tensorflow_core\python\saved_model\loader_impl.py in parse_saved_model(export_dir)
     81                   (export_dir,
     82                    constants.SAVED_MODEL_FILENAME_PBTXT,
---> 83                    constants.SAVED_MODEL_FILENAME_PB))
     84 
     85 

OSError: SavedModel file does not exist at: C:\Users\Asus\.keras\datasets\ssd_mobilenet_v1_coco_11_06_2017\saved_model/{saved_model.pbtxt|saved_model.pb}

Answer 1

我假设您在这里运行detection_model_zoo tutorial。请注意，也许您可以将型号名称从ssd_mobilenet_v1_coco_11_06_2017更改为ssd_mobilenet_v1_coco_2017_11_17，这将解决我的测试中的问题。

这些文件的内容可以在下面看到：

# ssd_mobilenet_v1_coco_11_06_2017
frozen_inference_graph.pb  model.ckpt.data-00000-of-00001  model.ckpt.meta
graph.pbtxt        model.ckpt.index

# ssd_mobilenet_v1_coco_2017_11_17
checkpoint         model.ckpt.data-00000-of-00001  model.ckpt.meta
frozen_inference_graph.pb  model.ckpt.index        saved_model

参考：

Answer 2

就我而言，此代码对我有用。我给出了由模型检查点模块创建的 .pd 文件的文件夹路径：

import tensorflow as tf
if __name__ == '__main__':
    # Update the input name and path for your Keras model
    input_keras_model = 'my path/weights/my_trained_model/{the files inside this folder are: assets(folder), variables(folder),keras_metadata.pd,saved_model.pd}'


    model = tf.keras.models.load_model(input_keras_model)

SavedModel文件在以下位置不存在：C：\ Users \ Asus \ .keras \ datasets \ ssd_mobilenet_v1_coco_11_06_2017 \ saved_model / {saved_model.pbtxt | saved_model.pb}

2 个答案: