Question

在TensorFlow的对象检测API中，可以在配置文件中指定一个keras模型（例如'ssd_mobilenet_v2_keras'），这样，构建的feature extractor和box predictor将是tf。 keras模型。我想做的就是使用这些创建一个简单的OD Keras模型。这意味着我将能够定义特征提取器，框预测器，然后执行以下操作：

def keras_model(model, shape):
    input = tf.keras.Input(shape=shape)
    feature_maps = model._feature_extractor(input)
    predictions = model._box_predictor(feature_maps)
    return tf.keras.Model(inputs=input, outputs=predictions)

使用model构建model_builder.build(model_config)时（将调用this）。

但是，当我尝试这样做时，在功能提取命令上出现以下异常：

    feature_maps = model._feature_extractor(input)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/engine/base_layer.py", line 584, in __call__
    inputs, outputs, args, kwargs)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/engine/base_layer.py", line 1416, in _set_connectivity_metadata_
    input_tensors=inputs, output_tensors=outputs, arguments=kwargs)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/engine/base_layer.py", line 1524, in _add_inbound_node
    arguments=arguments)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/engine/base_layer.py", line 1732, in __init__
    self.output_shapes = [backend.int_shape(x) for x in output_tensors]
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/engine/base_layer.py", line 1732, in <listcomp>
    self.output_shapes = [backend.int_shape(x) for x in output_tensors]
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/backend.py", line 951, in int_shape
    shape = x.shape
AttributeError: 'odict_values' object has no attribute 'shape'

任何想法是什么问题，或者是否有更简单的方法来提取对象检测Keras模型？

谢谢！

如何从对象检测API内置的模型中提取简单的对象检测Keras模型

0 个答案: