构建训练frcnn的模型，AttributeError classifier_layer

Question

我正在构建模型来训练frcnn卷积神经网络，问题是在建立模型时，它说我有一个错误，我怀疑这可能是由于tensorflow版本的不兼容。

下面的代码(错误第4行)：

# define the RPN, built on the base layers
num_anchors = len(C.anchor_box_scales) * len(C.anchor_box_ratios) # 9
rpn = rpn_layer(shared_layers, num_anchors)

classifier = classifier_layer(shared_layers, roi_input, C.num_rois, nb_classes=len(classes_count))

model_rpn = Model(img_input, rpn[:2])
model_classifier = Model([img_input, roi_input], classifier)

# this is a model that holds both the RPN and the classifier, used to load/save weights for the models
model_all = Model([img_input, roi_input], rpn[:2] + classifier)

# Because the google colab can only run the session several hours one time (then you need to connect again), 
# we need to save the model and load the model to continue training
if not os.path.isfile(C.model_path):
    #If this is the begin of the training, load the pre-traind base network such as vgg-16
    try:
        print('This is the first time of your training')
        print('loading weights from {}'.format(C.base_net_weights))
        model_rpn.load_weights(C.base_net_weights, by_name=True)
        model_classifier.load_weights(C.base_net_weights, by_name=True)
    except:
        print('Could not load pretrained model weights. Weights can be found in the keras application folder \
            https://github.com/fchollet/keras/tree/master/keras/applications')

    # Create the record.csv file to record losses, acc and mAP
    record_df = pd.DataFrame(columns=['mean_overlapping_bboxes', 'class_acc', 'loss_rpn_cls', 'loss_rpn_regr', 'loss_class_cls', 'loss_class_regr', 'curr_loss', 'elapsed_time', 'mAP'])
else:
    # If this is a continued training, load the trained model from before
    print('Continue training based on previous trained model')
    print('Loading weights from {}'.format(C.model_path))
    model_rpn.load_weights(C.model_path, by_name=True)
    model_classifier.load_weights(C.model_path, by_name=True)

    # Load the records
    record_df = pd.read_csv(record_path)

    r_mean_overlapping_bboxes = record_df['mean_overlapping_bboxes']
    r_class_acc = record_df['class_acc']
    r_loss_rpn_cls = record_df['loss_rpn_cls']
    r_loss_rpn_regr = record_df['loss_rpn_regr']
    r_loss_class_cls = record_df['loss_class_cls']
    r_loss_class_regr = record_df['loss_class_regr']
    r_curr_loss = record_df['curr_loss']
    r_elapsed_time = record_df['elapsed_time']
    r_mAP = record_df['mAP']

    print('Already train %dK batches'% (len(record_df)))

这就是输出：

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-55-039b2b159de7> in <module>()
      2 rpn = rpn_layer(shared_layers, num_anchors)
      3 
----> 4 classifier = classifier_layer(shared_layers, roi_input, C.num_rois, nb_classes=len(classes_count))
      5 
      6 model_rpn = Model(img_input, rpn[:2])

1 frames
<ipython-input-33-3ad8e8cb6f84> in __init__(self, pool_size, num_rois, **kwargs)
     19     def __init__(self, pool_size, num_rois, **kwargs):
     20 
---> 21         self.dim_ordering = K.image_dim_ordering()
     22         self.pool_size = pool_size
     23         self.num_rois = num_rois

AttributeError: module 'keras.backend' has no attribute 'image_dim_ordering'

我不知道怎么修复它，任何帮助我都很感激。

Answer 1

你使用的是什么版本的keras和TensorFlow？

尝试使用：

keras.backend.image_data_format()

Answer 2

如果你只想执行现有的代码，我建议将版本降级到keras 2.2.4和TF 1.13.1，这样你就可以执行它了。