farhodfm / keract

Activation Maps (Layers Outputs) and Gradients in Keras.

Keract: Keras Activations + Gradients

pip install keract

You have just found a (easy) way to get the activations (outputs) and gradients for each layer of your Keras model (LSTM, conv nets...).

API

• get_activations
• display_activations
• display_heatmaps
• get_gradients_of_trainable_weights
• get_gradients_of_activations
• persist_to_json_file
• load_activations_from_json_file

Get activations (outputs of each layer)

from keract import get_activations
activations = get_activations(model, x, layer_name)

Inputs are:

• model is a keras.models.Model object.
• x is a numpy array to feed to the model as input. In the case of multi-input, x is of type List. We use the Keras convention (as used in predict, fit...).
• layer_name (optional) - the layer to get activations for if you only want the activations for one layer

The output is a dictionary containing the activations for each layer of model for the input x:

{
  'conv2d_1/Relu:0': np.array(...),
  'conv2d_2/Relu:0': np.array(...),
  ...,
  'dense_2/Softmax:0': np.array(...)
}

The key is the name of the layer and the value is the corresponding output of the layer for the given input x.

Display the activations you've obtained

from keract import display_activations
display_activations(activations, cmap="gray", save=False)

Inputs are:

• activations a dictionary mapping layers to their activations (the output of get_activations)
• cmap (optional) a string of a valid matplotlib colourmap
• save(optional) a bool, if True the images of the activations are saved rather than being shown

Display the activations as a heatmap overlaid on an image

from keract import display_heatmaps
display_heatmaps(activations, input_image, save=False)

Inputs are:

• activations a dictionary mapping layers to their activations (the output of get_activations)
• input_image a numpy array of the image you inputed to the get_activations
• save(optional) a bool, if True the images of the activations are saved rather than being shown

Get gradients of weights

• model is a keras.models.Model object.
• x: Input data (numpy array). Keras convention.
• y: Labels (numpy array). Keras convention.

from keract import get_gradients_of_trainable_weights
get_gradients_of_trainable_weights(model, x, y)

The output is a dictionary mapping each trainable weight to the values of its gradients (regarding x and y).

Get gradients of activations

• model is a keras.models.Model object.
• x: Input data (numpy array). Keras convention.
• y: Labels (numpy array). Keras convention.

from keract import get_gradients_of_activations
get_gradients_of_activations(model, x, y)

The output is a dictionary mapping each layer to the values of its gradients (regarding x and y).

Persist activations to JSON

• activations: activations (dict mapping layers)
• filename: output filename (JSON format)

from keract import persist_to_json_file
persist_to_json_file(activations, filename)

Load activations from JSON

• filename: filename to read the activations from (JSON format)

from keract import persist_to_json_file
load_activations_from_json_file(filename)

It returns the activations.

Examples

Examples are provided for:

• keras.models.Sequential - mnist.py
• keras.models.Model - multi_inputs.py
• Recurrent networks - recurrent.py

In the case of MNIST with LeNet, we are able to fetch the activations for a batch of size 128:

conv2d_1/Relu:0
(128, 26, 26, 32)

conv2d_2/Relu:0
(128, 24, 24, 64)

max_pooling2d_1/MaxPool:0
(128, 12, 12, 64)

dropout_1/cond/Merge:0
(128, 12, 12, 64)

flatten_1/Reshape:0
(128, 9216)

dense_1/Relu:0
(128, 128)

dropout_2/cond/Merge:0
(128, 128)

dense_2/Softmax:0
(128, 10)

We can visualise the activations. Here's another example using VGG16:

cd examples
pip install -r examples-requirements.txt
python vgg16.py

A cat.

Outputs of the first convolutional layer of VGG16.

Also, we can visualise the heatmaps of the activations:

cd examples
pip install -r examples-requirements.txt
python heat_map.py

Tests

Testing is based on Tox.

pip install tox
tox