Issue with tensorflow/ keras densenet

TensorFlow General Discussion
, ,
1

Hi everybody,

When I am running the code pasted below, the model is just training for “multiplier” =1 or =4.
Running the same code in google colab → just training for multiplier=1

Is there any mistake in how I am using DenseNet here?

Thanks in advance, appreciate your help!

import numpy as np
import tensorflow as tf
from tensorflow.keras.applications.densenet import DenseNet201
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.losses import BinaryCrossentropy


random_array = np.random.rand(128,128,3)
image = tf.convert_to_tensor(
    random_array
)
label = tf.constant(0)



model = DenseNet201(
    include_top=False, weights='imagenet', input_tensor=None,
    input_shape=(128, 128, 3), pooling=None, classes=2
)
model.compile(
optimizer=Adam(),
loss=BinaryCrossentropy(),
metrics=['accuracy'],
)


for multiplier in range(1,20):

    print(f"Using multiplier {multiplier}")
    x_train = np.array([image]*multiplier)
    y_train = np.array([label]*multiplier)



    try: 
        model.fit(x=x_train,y=y_train, epochs=2)
    except:
        print("Not training...")
        pass
3

Some advice:

  • Test small pieces before trying to put them together.
  • Read your error messages.

The problem is that you set “include_top=False”:

model = DenseNet201(
  include_top=False, weights='imagenet', input_tensor=None,
  input_shape=(128, 128, 3), pooling=None, classes=2
)

So the output is a batch of feature maps:

result = model(x_train)
result.shape

TensorShape([2, 4, 4, 1920])

And you’ve set the label to [0,0] and the loss function to BinaryCrossentropy.

model.compile(
  optimizer=Adam(),
  loss=BinaryCrossentropy(),
  metrics=['accuracy'],
  run_eagerly=True,
)

It’s failing because bce=BinaryCrossentropy(); bce([0,0], result) doesn’t work.

That’s what the error message was trying to tell you.

5

Thanks Mark!

From the documentation on the preptrained models, it wasn’t clear to me, that it is necessary to specify your own Classification Layer. I thought by defining the number of classes, it would be added automatically.