Feature map visualization in TensorFlow - ML Experiment: Train & Evaluate

Experiment - Feature map visualization

Problem:You have trained a convolutional neural network (CNN) on the MNIST dataset to classify handwritten digits. You want to understand what the CNN learns by visualizing the feature maps (activation outputs) of the first convolutional layer for a sample input image.

Current Metrics:Training accuracy: 98%, Validation accuracy: 97%

Issue:Although the model performs well, you cannot see what features the CNN extracts from the input images. You want to visualize the feature maps to gain insight into the learned filters.

Your Task

Visualize the feature maps produced by the first convolutional layer of the trained CNN for a given input image from the test set.

Use TensorFlow and Keras only.

Do not retrain the model; use the existing trained model.

Visualize feature maps as grayscale images in a grid.

Hint 1

Hint 2

Hint 3

Solution

TensorFlow

import tensorflow as tf
from tensorflow.keras import layers, models
import matplotlib.pyplot as plt
import numpy as np

# Load MNIST dataset
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()

# Normalize and reshape data
x_train = x_train.astype('float32') / 255.0
x_test = x_test.astype('float32') / 255.0
x_train = np.expand_dims(x_train, -1)  # shape (num_samples, 28, 28, 1)
x_test = np.expand_dims(x_test, -1)

# Define a simple CNN model
model = models.Sequential([
    layers.Conv2D(32, (3,3), activation='relu', input_shape=(28,28,1)),
    layers.MaxPooling2D((2,2)),
    layers.Conv2D(64, (3,3), activation='relu'),
    layers.MaxPooling2D((2,2)),
    layers.Flatten(),
    layers.Dense(64, activation='relu'),
    layers.Dense(10, activation='softmax')
])

# Compile and train the model
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
model.fit(x_train, y_train, epochs=3, batch_size=64, validation_split=0.1, verbose=0)

# Select one test image
test_img = x_test[0]

# Create a model to output the feature maps of the first Conv2D layer
layer_outputs = [layer.output for layer in model.layers if isinstance(layer, layers.Conv2D)]
first_conv_layer_model = models.Model(inputs=model.input, outputs=layer_outputs[0])

# Get feature maps for the test image
feature_maps = first_conv_layer_model.predict(np.expand_dims(test_img, axis=0))  # shape (1, 26, 26, 32)

# Plot the feature maps
num_filters = feature_maps.shape[-1]
size = feature_maps.shape[1]

cols = 8
rows = num_filters // cols
plt.figure(figsize=(cols*1.5, rows*1.5))
for i in range(num_filters):
    plt.subplot(rows, cols, i+1)
    plt.imshow(feature_maps[0, :, :, i], cmap='gray')
    plt.axis('off')
plt.suptitle('Feature maps of first Conv2D layer')
plt.show()

Created a new model that outputs the activations of the first convolutional layer.

Selected one test image and preprocessed it to match model input shape.

Used matplotlib to plot each feature map as a grayscale image in a grid.

Results Interpretation

Before: Model accuracy was high but no insight into learned features.

After: Visualized 32 feature maps showing different patterns the first convolutional layer detects from the input image.

Visualizing feature maps helps understand what patterns a CNN learns at different layers, improving interpretability without changing model performance.

Bonus Experiment

Visualize feature maps from the second convolutional layer and compare them with the first layer's feature maps.

💡 Hint

Create a similar model outputting the second Conv2D layer activations and plot them. Notice how deeper layers capture more complex features.

Practice

1. What is the main purpose of visualizing feature maps in a convolutional neural network (CNN)?

easy

A. To reduce the size of the input images

B. To understand what patterns or features the CNN layers detect in the input image

C. To increase the training speed of the CNN model

D. To change the activation function of the CNN layers

Solution

1. Step 1: Understand feature maps in CNNs

   Feature maps are outputs of convolutional layers showing detected features like edges or textures.

2. Step 2: Purpose of visualization

   Visualizing feature maps helps us see what the CNN focuses on in the image, improving model interpretability.

3. Final Answer:

   To understand what patterns or features the CNN layers detect in the input image -> Option B

4. Quick Check:

   Feature map visualization = Understand CNN features [OK]

Hint: Feature maps show what CNN 'sees' in images [OK]

Common Mistakes:

• Thinking visualization speeds up training
• Confusing feature maps with input resizing
• Believing visualization changes activation functions

2. Which TensorFlow code snippet correctly creates a model to output feature maps from the first convolutional layer named 'conv1'?

easy

A. feature_map_model = tf.keras.Model(inputs=model.get_layer('conv1').input, outputs=model.input)

B. feature_map_model = tf.keras.Sequential(model.get_layer('conv1'))

C. feature_map_model = tf.keras.Model(inputs=model.input, outputs=model.get_layer('conv1').output)

D. feature_map_model = tf.keras.Model(inputs=model.output, outputs=model.get_layer('conv1').input)

Solution

1. Step 1: Identify correct Model creation syntax

   To get intermediate outputs, create a new Model with original inputs and desired layer outputs.

2. Step 2: Check each option

   feature_map_model = tf.keras.Model(inputs=model.input, outputs=model.get_layer('conv1').output) correctly uses model.input and conv1 output. Others misuse inputs/outputs or use Sequential incorrectly.

3. Final Answer:

   feature_map_model = tf.keras.Model(inputs=model.input, outputs=model.get_layer('conv1').output) -> Option C

4. Quick Check:

   New model inputs = original inputs, outputs = layer output [OK]

Hint: Use original inputs and layer output to create feature map model [OK]

Common Mistakes:

• Using model.output as input
• Passing layer input as output
• Using Sequential instead of Model for intermediate outputs

3. Given the code below, what is the shape of feature_maps if the input image shape is (1, 64, 64, 3) and the 'conv1' layer outputs 32 filters of size 62x62?

feature_map_model = tf.keras.Model(inputs=model.input, outputs=model.get_layer('conv1').output)
feature_maps = feature_map_model.predict(input_image)

medium

A. (62, 62, 32)

B. (32, 62, 62)

C. (1, 64, 64, 3)

D. (1, 62, 62, 32)

Solution

1. Step 1: Understand input and output shapes

   Input batch size is 1, image size 64x64 with 3 channels. conv1 outputs 32 filters of size 62x62.

2. Step 2: Predict output shape

   Output shape includes batch size, height, width, and number of filters: (1, 62, 62, 32).

3. Final Answer:

   (1, 62, 62, 32) -> Option D

4. Quick Check:

   Output shape = (batch, height, width, filters) [OK]

Hint: Feature map shape includes batch and filter count [OK]

Common Mistakes:

• Ignoring batch dimension
• Confusing input shape with output shape
• Mixing filter count with spatial dimensions

4. You wrote this code to visualize feature maps but get an error:

feature_map_model = tf.keras.Model(inputs=model.input, outputs=model.get_layer('conv1').output)
feature_maps = feature_map_model.predict(input_image)
plt.imshow(feature_maps)

What is the likely cause of the error?

medium

A. Trying to plot a 4D tensor directly with plt.imshow

B. Using the wrong layer name in get_layer

C. Input image shape is incompatible with the model

D. Missing import for matplotlib.pyplot

Solution

1. Step 1: Analyze the error cause

   feature_maps is a 4D array (batch, height, width, filters), but plt.imshow expects 2D or 3D image data.

2. Step 2: Correct visualization approach

   You must select one feature map (e.g., feature_maps[0, :, :, 0]) before plotting with plt.imshow.

3. Final Answer:

   Trying to plot a 4D tensor directly with plt.imshow -> Option A

4. Quick Check:

   Plot single 2D feature map slice, not full 4D tensor [OK]

Hint: Plot one 2D feature map slice, not whole 4D tensor [OK]

Common Mistakes:

• Plotting entire feature_maps array directly
• Ignoring batch dimension when plotting
• Assuming get_layer error without checking data shape

5. You want to visualize feature maps from multiple convolutional layers ('conv1', 'conv2', 'conv3') in a TensorFlow model. Which approach correctly creates a model to output all these feature maps at once?

hard

A. feature_map_model = tf.keras.Model(inputs=model.input, outputs=[model.get_layer(name).output for name in ['conv1', 'conv2', 'conv3']])

B. feature_map_model = tf.keras.Sequential([model.get_layer('conv1'), model.get_layer('conv2'), model.get_layer('conv3')])

C. feature_map_model = tf.keras.Model(inputs=model.output, outputs=model.get_layer('conv3').output)

D. feature_map_model = tf.keras.Model(inputs=model.input, outputs=model.get_layer('conv3').input)

Solution

1. Step 1: Understand multi-output model creation

   To get outputs from multiple layers, pass a list of their outputs to the Model constructor.

2. Step 2: Evaluate options

   feature_map_model = tf.keras.Model(inputs=model.input, outputs=[model.get_layer(name).output for name in ['conv1', 'conv2', 'conv3']]) correctly creates a model with inputs as original inputs and outputs as a list of layer outputs. Others misuse Sequential or inputs/outputs.

3. Final Answer:

   feature_map_model = tf.keras.Model(inputs=model.input, outputs=[model.get_layer(name).output for name in ['conv1', 'conv2', 'conv3']]) -> Option A

4. Quick Check:

   Use list of outputs for multiple feature maps [OK]

Hint: Use list of layer outputs in Model to get multiple feature maps [OK]

Common Mistakes:

• Using Sequential instead of Model for intermediate outputs
• Passing model.output as inputs
• Using layer inputs as outputs