AttributeError: 'Model' object has no attribute 'metrics_tensors' running mask r-cnn

[jiangjiangjianggx]

Hi, I am trying to run a mask r-cnn code for dental segmentation images training based on the coco and Mask_RCNN, the code should work perfectly, but since my Keras is 2.3.0 and it seems not having the attribute metrics_tensor.

the code:

model.train(dataset_train, dataset_train,
            learning_rate=config_train.LEARNING_RATE,
            epochs=10,
            layers='heads')

# Training - Stage 2
# Finetune layers from ResNet stage 4 and up
print("Fine tune Resnet stage 4 and up")
model.train(dataset_train, dataset_train,
            learning_rate=config_train.LEARNING_RATE,
            epochs=50,
            layers='4+')

# Training - Stage 3
# Fine tune all layers
print("Fine tune all layers")
model.train(dataset_train, dataset_train,
            learning_rate=config_train.LEARNING_RATE / 10,
            epochs=100,
            layers='all'

here is the error:

AttributeError Traceback (most recent call last) in 4 learning_rate=config_train.LEARNING_RATE, 5 epochs=10, ----> 6 layers='heads') 7 8 # Training - Stage 2

/usr/local/lib/python3.7/site-packages/mrcnn/model.py in train(self, train_dataset, val_dataset, learning_rate, epochs, layers, augmentation) 2330 log("Checkpoint Path: {}".format(self.checkpoint_path)) 2331 self.set_trainable(layers) -> 2332 self.compile(learning_rate, self.config.LEARNING_MOMENTUM) 2333 2334 # Work-around for Windows: Keras fails on Windows when using

/usr/local/lib/python3.7/site-packages/mrcnn/model.py in compile(self, learning_rate, momentum) 2186 tf.reduce_mean(layer.output, keepdims=True) 2187 * self.config.LOSS_WEIGHTS.get(name, 1.)) -> 2188 self.keras_model.metrics_tensors.append(loss) 2189 2190 def set_trainable(self, layer_regex, keras_model=None, indent=0, verbose=1):

AttributeError: 'Model' object has no attribute 'metrics_tensors'

how can I fix this error? should I downgrade my Keras to lower version?

[atahadd]

I have the same issue with keras=1.3.1

[DataStunner]

It is metrics instead of metrics_tensors. I'm using Tensorflow 1.14.0 and Keras 2.3.0. Updating the model.py (2190 - 2199) worked for me.

Add metrics for losses

    for name in loss_names:
        if name in self.keras_model.metrics_names:
            continue
        layer = self.keras_model.get_layer(name)
        self.keras_model.metrics_names.append(name)
        loss = (
            tf.reduce_mean(layer.output, keepdims=True)
            * self.config.LOSS_WEIGHTS.get(name, 1.))
        self.keras_model.metrics.append(loss)

[dlllll-q]

It is metrics instead of metrics_tensors. I'm using Tensorflow 1.14.0 and Keras 2.3.0. Updating the model.py (2190 - 2199) worked for me.

Add metrics for losses

    for name in loss_names:
        if name in self.keras_model.metrics_names:
            continue
        layer = self.keras_model.get_layer(name)
        self.keras_model.metrics_names.append(name)
        loss = (
            tf.reduce_mean(layer.output, keepdims=True)
            * self.config.LOSS_WEIGHTS.get(name, 1.))
        self.keras_model.metrics.append(loss)

Hi, firstly, thanks for your reply, this works for me. But there is one problem, in tf.1.12, I use metrics_tensor, which can print loc_loss, class_loss except the total loss. When i change metrics_tensor to metrics in tf1.14, only total loss can be printed in training process. How can I solve this problem?

[pipun-gif]

It is metrics instead of metrics_tensors. I'm using Tensorflow 1.14.0 and Keras 2.3.0. Updating the model.py (2190 - 2199) worked for me.

Add metrics for losses

    for name in loss_names:
        if name in self.keras_model.metrics_names:
            continue
        layer = self.keras_model.get_layer(name)
        self.keras_model.metrics_names.append(name)
        loss = (
            tf.reduce_mean(layer.output, keepdims=True)
            * self.config.LOSS_WEIGHTS.get(name, 1.))
        self.keras_model.metrics.append(loss)

Hi, firstly, thanks for your reply, this works for me. But there is one problem, in tf.1.12, I use metrics_tensor, which can print loc_loss, class_loss except the total loss. When i change metrics_tensor to metrics in tf1.14, only total loss can be printed in training process. How can I solve this problem?

hi excuse me @dlllll-q how to edit model.py inside egg file? i can open it by rename the .egg into .zip and modify all file inside that egg, but unfortunately i don't know package it back as a python egg file after my modification. thank you

[rsalunga29]

It is metrics instead of metrics_tensors. I'm using Tensorflow 1.14.0 and Keras 2.3.0. Updating the model.py (2190 - 2199) worked for me.

Add metrics for losses

    for name in loss_names:
        if name in self.keras_model.metrics_names:
            continue
        layer = self.keras_model.get_layer(name)
        self.keras_model.metrics_names.append(name)
        loss = (
            tf.reduce_mean(layer.output, keepdims=True)
            * self.config.LOSS_WEIGHTS.get(name, 1.))
        self.keras_model.metrics.append(loss)

do we have a pull request for this so it can be fixed for everyone running keras 2.3.x?

[rrm003]

It is metrics instead of metrics_tensors. I'm using Tensorflow 1.14.0 and Keras 2.3.0. Updating the model.py (2190 - 2199) worked for me.

Add metrics for losses

    for name in loss_names:
        if name in self.keras_model.metrics_names:
            continue
        layer = self.keras_model.get_layer(name)
        self.keras_model.metrics_names.append(name)
        loss = (
            tf.reduce_mean(layer.output, keepdims=True)
            * self.config.LOSS_WEIGHTS.get(name, 1.))
        self.keras_model.metrics.append(loss)

I am getting AttributeError: 'NoneType' object has no attribute 'append' for self.keras_model.metrics.append(loss)

[weishanlee]

It is metrics instead of metrics_tensors. I'm using Tensorflow 1.14.0 and Keras 2.3.0. Updating the model.py (2190 - 2199) worked for me.

Add metrics for losses

    for name in loss_names:
        if name in self.keras_model.metrics_names:
            continue
        layer = self.keras_model.get_layer(name)
        self.keras_model.metrics_names.append(name)
        loss = (
            tf.reduce_mean(layer.output, keepdims=True)
            * self.config.LOSS_WEIGHTS.get(name, 1.))
        self.keras_model.metrics.append(loss)

I am getting AttributeError: 'NoneType' object has no attribute 'append' for self.keras_model.metrics.append(loss)

Try to restart all over again to run the script. It should be ok.