keras - 值错误 : Found two metrics with the same name: recall

Question

我正在训练一个检测模型，其中训练和测试数据是 3D NumPy 数组。当开始训练这个模型时发现了这种类型的错误。代码链接如下

Training_model.py

detection.py

perform_learning.py

model.fit_generator(generator=training_generator,
                        validation_data=validation_generator,
                        use_multiprocessing=True,
                        workers=6,
                        epochs=epochs,
                        callbacks=[checkpoint, tensorboard])

追溯(最近的调用最后):文件“/content/SpineFinder-master/train_detection_model.py”，第 25 行，位于洗牌=真)

文件“/content/SpineFinder-master/learning_functions/perform_learning.py”，第 57 行，在 perform_learning回调=[检查点，张量板])

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/util/deprecation.py”，第 324 行，在 new_func返回 func(*args, **kwargs)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py”，第 1479 行，在 fit_generator 中initial_epoch=initial_epoch)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py”，第 66 行，在 _method_wrapper 中返回方法(self, *args, **kwargs)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py”，第 848 行，适合tmp_logs = train_function(迭代器)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/def_function.py”，第 580 行，在 call结果 = self._call(*args, **kwds)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/def_function.py”，第 627 行，在 _call 中self._initialize(args, kwds, add_initializers_to=initializers)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/def_function.py”，第 506 行，在 _initialize 中*args, **kwds))

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/function.py”，第 2446 行，在 _get_concrete_function_internal_garbage_collectedgraph_function,_,_=self._maybe_define_function(args, kwargs)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/function.py”，第 2777 行，在 _maybe_define_function 中graph_function = self._create_graph_function(args, kwargs)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/function.py”，第 2667 行，在 _create_graph_function 中capture_by_value=self._capture_by_value),

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/func_graph.py”，第 981 行，在 func_graph_from_py_funcfunc_outputs = python_func(*func_args, **func_kwargs)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/def_function.py”，第 441 行，wrapped_fn返回 weak_wrapped_fn()。包装(*args, **kwds)

文件“/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/func_graph.py”，第 968 行，在包装器中提高 e.ag_error_metadata.to_exception(e)

ValueError: 在用户代码中:

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:571 train_function *outputs = self.distribute_strategy.run(/usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:951 运行 **返回 self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:2290 call_for_each_replica
    return self._call_for_each_replica(fn, args, kwargs)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:2649 _call_for_each_replica
    return fn(*args, **kwargs)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:543 train_step  **
    self.compiled_metrics.update_state(y, y_pred, sample_weight)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/compile_utils.py:391 update_state
    self._build(y_pred, y_true)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/compile_utils.py:333 _build
    self._set_metric_names()

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/compile_utils.py:353 _set_metric_names
    m._name))

ValueError:找到两个同名的指标:recall

Answer 1

错误源于 detection.py 中的以下内容:

    recall_background = km.binary_recall(label=0)
    recall_vertebrae = km.binary_recall(label=1)

根据[1]和[2]，km.binary_recall()实例化keras.metrics.recall()类。但是，如果没有 name kwarg，两行都使用相同的名称 recall。因此，为了避免这种情况，我的理解是您必须像这样指定 name kwarg:

    recall_background = km.binary_recall(name="recall_background", label=0)
    recall_vertebrae = km.binary_recall(name="recall_vertebrae", label=1)

[1] - https://github.com/netrack/keras-metrics/blob/master/keras_metrics/\_\_init__.py#L34

[2] - https://github.com/netrack/keras-metrics/blob/master/keras_metrics/metrics.py#L150