Блог

Я пытаюсь обучить сеть регрессии с ограничением. Я создал pd.DataFrame, который выглядит следующим образом:

Вот мои генераторы обучающих и проверочных изображений:

 image_generator = tf.keras.preprocessing.image.ImageDataGenerator(
                                  rescale = 1./255,
                                  rotation_range = 25,
                                  zoom_range=[0.8, 1.2],
                                  vertical_flip=True,
                                  horizontal_flip=True,
                                  )
train_generator = image_generator.flow_from_dataframe(
            dataframe=train_df,
            directory=cbis_ddsm_train_images_dir,
            x_col="image file path",
            y_col="coordinates",
            class_mode="raw",
            batch_size=BATCH_SIZE,
            shuffle=True,
            seed=1,
            color_mode="grayscale",
            target_size=(1024, 1024))

val_gen = tf.keras.preprocessing.image.ImageDataGenerator(
                                  rescale = 1./255,
                                  )

val_generator = val_gen.flow_from_dataframe(
            dataframe=val_df,
            directory=cbis_ddsm_train_images_dir,
            x_col="image file path",
            y_col=None,
            class_mode="raw",
            batch_size=BATCH_SIZE,
            shuffle=False,
            seed=1,
            color_mode="grayscale",
             target_size=(1024, 1024))
 

Обратите внимание, что я преобразовал столбец Y (т. Е. Координаты bbox) из списков в числовой размерный массив следующим образом:

 for idx, row in train_df.iterrows():
  height, width = row['size']
  row['coordinates'] = np.asarray([normalize_bbox(c, height, width) for c in row['coordinates']][0]).astype('float32')
 

Когда я пытаюсь выполнить код ниже:

 opt = Adam(lr=INIT_LR)
final_model.compile(optimizer=opt, loss="mse")
    
final_model.fit(train_generator, steps_per_epoch=steps_per_epoch, epochs=3,
                    validation_data=val_generator, validation_steps=val_steps, verbose=1)
 

Я получаю следующее сообщение об ошибке:

 ValueError                                Traceback (most recent call last)
<ipython-input-46-90f3a1cd9c87> in <module>()
      8 
      9     final_model.fit(train_generator, steps_per_epoch=steps_per_epoch, epochs=3,
---> 10                     validation_data=val_generator, validation_steps=val_steps, verbose=1)
     11     final_model.save(os.path.join(experiment1_dir, "resnet_fine-tuned-head.h5"))
     12 

14 frames
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py in _method_wrapper(self, *args, **kwargs)
    106   def _method_wrapper(self, *args, **kwargs):
    107     if not self._in_multi_worker_mode():  # pylint: disable=protected-access
--> 108       return method(self, *args, **kwargs)
    109 
    110     # Running inside `run_distribute_coordinator` already.

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, validation_batch_size, validation_freq, max_queue_size, workers, use_multiprocessing)
   1061           use_multiprocessing=use_multiprocessing,
   1062           model=self,
-> 1063           steps_per_execution=self._steps_per_execution)
   1064 
   1065       # Container that configures and calls `tf.keras.Callback`s.

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in __init__(self, x, y, sample_weight, batch_size, steps_per_epoch, initial_epoch, epochs, shuffle, class_weight, max_queue_size, workers, use_multiprocessing, model, steps_per_execution)
   1115         use_multiprocessing=use_multiprocessing,
   1116         distribution_strategy=ds_context.get_strategy(),
-> 1117         model=model)
   1118 
   1119     strategy = ds_context.get_strategy()

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in __init__(self, x, y, sample_weights, shuffle, workers, use_multiprocessing, max_queue_size, model, **kwargs)
    914         max_queue_size=max_queue_size,
    915         model=model,
--> 916         **kwargs)
    917 
    918   @staticmethod

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in __init__(self, x, y, sample_weights, workers, use_multiprocessing, max_queue_size, model, **kwargs)
    786     peek, x = self._peek_and_restore(x)
    787     peek = self._standardize_batch(peek)
--> 788     peek = _process_tensorlike(peek)
    789 
    790     # Need to build the Model on concrete input shapes.

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in _process_tensorlike(inputs)
   1019     return x
   1020 
-> 1021   inputs = nest.map_structure(_convert_numpy_and_scipy, inputs)
   1022   return nest.list_to_tuple(inputs)
   1023 

/usr/local/lib/python3.6/dist-packages/tensorflow/python/util/nest.py in map_structure(func, *structure, **kwargs)
    633 
    634   return pack_sequence_as(
--> 635       structure[0], [func(*x) for x in entries],
    636       expand_composites=expand_composites)
    637 

/usr/local/lib/python3.6/dist-packages/tensorflow/python/util/nest.py in <listcomp>(.0)
    633 
    634   return pack_sequence_as(
--> 635       structure[0], [func(*x) for x in entries],
    636       expand_composites=expand_composites)
    637 

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in _convert_numpy_and_scipy(x)
   1014       if issubclass(x.dtype.type, np.floating):
   1015         dtype = backend.floatx()
-> 1016       return ops.convert_to_tensor(x, dtype=dtype)
   1017     elif scipy_sparse and scipy_sparse.issparse(x):
   1018       return _scipy_sparse_to_sparse_tensor(x)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/ops.py in convert_to_tensor(value, dtype, name, as_ref, preferred_dtype, dtype_hint, ctx, accepted_result_types)
   1497 
   1498     if ret is None:
-> 1499       ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
   1500 
   1501     if ret is NotImplemented:

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/tensor_conversion_registry.py in _default_conversion_function(***failed resolving arguments***)
     50 def _default_conversion_function(value, dtype, name, as_ref):
     51   del as_ref  # Unused.
---> 52   return constant_op.constant(value, dtype, name=name)
     53 
     54 

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/constant_op.py in constant(value, dtype, shape, name)
    262   """
    263   return _constant_impl(value, dtype, shape, name, verify_shape=False,
--> 264                         allow_broadcast=True)
    265 
    266 

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/constant_op.py in _constant_impl(value, dtype, shape, name, verify_shape, allow_broadcast)
    273       with trace.Trace("tf.constant"):
    274         return _constant_eager_impl(ctx, value, dtype, shape, verify_shape)
--> 275     return _constant_eager_impl(ctx, value, dtype, shape, verify_shape)
    276 
    277   g = ops.get_default_graph()

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/constant_op.py in _constant_eager_impl(ctx, value, dtype, shape, verify_shape)
    298 def _constant_eager_impl(ctx, value, dtype, shape, verify_shape):
    299   """Implementation of eager constant."""
--> 300   t = convert_to_eager_tensor(value, ctx, dtype)
    301   if shape is None:
    302     return t

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/constant_op.py in convert_to_eager_tensor(value, ctx, dtype)
     96       dtype = dtypes.as_dtype(dtype).as_datatype_enum
     97   ctx.ensure_initialized()
---> 98   return ops.EagerTensor(value, ctx.device_name, dtype)
     99 
    100 

ValueError: Failed to convert a NumPy array to a Tensor (Unsupported object type numpy.ndarray).
 

Любая помощь была бы ОЧЕНЬ признательна. Спасибо!

Это ошибка в Keras, о которой сообщается здесь: https://github.com/keras-team/keras/issues/13839

По сути, когда class_mode == "raw" и метки являются массивами numpy, flow_from_dataframe генерируются пакеты для меток в форме массива массивов numpy, а не 2D-массива, что затем приводит к сбою метода подгонки.

В качестве обходного пути, пока он не будет исправлен, добавьте эти строки после создания ваших генераторов

 train_generator._targets = np.stack(train_generator._targets)
val_generator._targets = np.stack(val_generator._targets)