[FIXED] How to save a neural network model in h5 with loss function "balanced categorical entropy"?

Issue

I am using VGG16 for image segmentation with the loss function "balanced categorical entropy" using the code

beta=0.5

def balanced_cross_entropy(beta):
  def loss(y_true, y_pred):
    weight_a = beta * tf.cast(y_true, tf.float32)
    weight_b = (1 - beta) * tf.cast(1 - y_true, tf.float32)
    
    o = (tf.math.log1p(tf.exp(-tf.abs(y_pred))) + tf.nn.relu(-y_pred)) * (weight_a + weight_b) + y_pred * weight_b
    return tf.reduce_mean(o)

  return loss

Everything works fine. Now I save this model in the h5 file using the code.

vgg.save('vgg.h5')

But when I use the load_model from Keras

 model = load_model('vgg.h5', custom_objects={'balanced_cross_entropy(beta)': balanced_cross_entropy(beta)})

I encounter an error.

Unknown loss function: loss. Please ensure this object is passed to the `custom_objects` argument.

Can anybody help, I suspect the problem may be due to beta?

Solution

If you want to only perform inference, you can avoid this problem by specifying

model = load_model('vgg.h5',compile=False)

Otherwise, you need to load the in the following way:

model = load_model("vgg.h5", custom_objects={'loss': balanced_cross_entropy(beta)}); in your code you wrote balanced_cross_entropy(beta) instead of loss.

Short explanation:

The name of the key in custom_object is actually the name of the inner function (which is in fact returned by balanced_cross_entropy(beta); the name of the outer function is actually the value of the <key,value> pair in the custom_object dictionary.

Answered By - Timbus Calin