CNN architecture for image classification in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output shape after Conv2D layer

Given the following TensorFlow code snippet for a CNN layer, what is the output shape of the tensor after the Conv2D layer?

TensorFlow

import tensorflow as tf
input_tensor = tf.random.uniform([1, 64, 64, 3])
conv_layer = tf.keras.layers.Conv2D(filters=16, kernel_size=3, strides=2, padding='same')
output_tensor = conv_layer(input_tensor)
print(output_tensor.shape)

A (1, 64, 64, 16)

B (1, 31, 31, 16)

C (1, 62, 62, 16)

D (1, 32, 32, 16)

💡 Hint

Remember that padding='same' keeps the spatial size divided by stride.

With padding='same' and stride=2, the output height and width are input size divided by stride, rounded up. So 64/2=32. The number of filters is 16, so output shape is (1, 32, 32, 16).

❓ Model Choice

intermediate

Choosing activation function for CNN output layer

You are building a CNN to classify images into 10 categories. Which activation function should you use in the output layer to get probabilities for each class?

A Softmax

B ReLU

C Sigmoid

D Tanh

💡 Hint

The output should represent probabilities that sum to 1.

Softmax converts raw scores into probabilities that sum to 1, which is ideal for multi-class classification.

❓ Hyperparameter

advanced

Effect of increasing kernel size in Conv2D

In a CNN, what is the most likely effect of increasing the kernel size from 3x3 to 7x7 in the first Conv2D layer?

A The model captures larger spatial features but increases parameters and computation

B The model reduces overfitting by decreasing parameters

C The model output size increases significantly

D The model training speed increases due to fewer computations

💡 Hint

Think about how kernel size affects receptive field and parameter count.

Larger kernels cover more area, capturing bigger features but also increasing the number of parameters and computation cost.

❓ Metrics

advanced

Interpreting CNN training accuracy and loss

During CNN training for image classification, you observe training accuracy steadily increasing but validation accuracy plateaus and then decreases. What does this indicate?

A The dataset is too small to train

B The model is underfitting the training data

C The model is overfitting the training data

D The learning rate is too low

💡 Hint

Think about what it means when training improves but validation worsens.

When training accuracy improves but validation accuracy drops, the model memorizes training data and fails to generalize, which is overfitting.

🔧 Debug

expert

Identifying error in CNN model definition

What error will this TensorFlow CNN model code raise when run?

TensorFlow

import tensorflow as tf
model = tf.keras.Sequential([
    tf.keras.layers.Conv2D(32, kernel_size=3, activation='relu', input_shape=(28, 28, 1)),
    tf.keras.layers.MaxPooling2D(pool_size=2),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(10, activation='softmax')
])
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])

A RuntimeError: Missing Flatten layer before Dense

B ValueError: Input shape must have 3 dimensions for Conv2D

C TypeError: activation function 'relu' not recognized

D No error, model compiles successfully

💡 Hint

Check the input_shape parameter for Conv2D layer.

Conv2D expects input shape with 3 dimensions: height, width, and channels. Here channels dimension is missing.