Feature map visualization in TensorFlow - Model Metrics & Evaluation

Metrics & Evaluation - Feature map visualization

Which metric matters for Feature map visualization and WHY

Feature map visualization helps us see what parts of an image a neural network focuses on. It is not about accuracy or loss numbers. Instead, it shows the patterns the model learns inside its layers. This helps us understand if the model looks at meaningful features or just noise. So, the key metric here is interpretability, not a number. We want clear, understandable feature maps that highlight important image parts.

Confusion matrix or equivalent visualization

Feature map visualization is about images, so instead of a confusion matrix, we use visual maps. For example, after a convolution layer, the feature map might look like a grid of small images showing edges, colors, or shapes the model detects.

Input Image --> Conv Layer --> Feature Maps (e.g., 16 maps of 28x28 pixels)
Each map highlights different features like edges or textures.

Example feature map grid:
[Map1] [Map2] [Map3] ... [Map16]

Precision vs Recall tradeoff with concrete examples

Feature map visualization is not about precision or recall. But there is a tradeoff in clarity vs detail. If feature maps are too detailed, they may be noisy and hard to interpret. If too simple, they may miss important features. For example, showing all 64 feature maps from a deep layer can be overwhelming, while showing only a few may hide useful info. The tradeoff is between too much info and too little.

What "good" vs "bad" metric values look like for this use case

Good feature maps clearly highlight meaningful parts of the input, like edges of objects or textures related to the task. They look structured and consistent across similar images.

Bad feature maps look noisy, random, or highlight irrelevant areas. They may be blurry or show no clear pattern, meaning the model might not be learning useful features.

Metrics pitfalls

• Confusing visualization with performance: Beautiful feature maps don't always mean a good model. The model might still perform poorly on real data.
• Overfitting signs: Feature maps that focus too narrowly on small details or noise may indicate overfitting.
• Data leakage: If feature maps highlight parts that leak label info (like watermarks), the model may cheat.
• Ignoring scale: Feature maps from early layers show simple features; expecting complex patterns there is wrong.

Self-check question

Your model shows clear feature maps highlighting object edges, but its accuracy is low. Is the model good? Why or why not?

Answer: Clear feature maps mean the model learns some useful features, but low accuracy shows it is not enough. The model might need better training or architecture. So, it is not good yet despite nice visualizations.

Key Result

Feature map visualization focuses on interpretability of learned patterns, not numeric metrics like accuracy.

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