CNN architecture review in Computer Vision - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is the main purpose of convolutional layers in a CNN?

Convolutional layers detect local patterns like edges or textures by sliding small filters over the input image, helping the model learn important features.

beginner

Why do CNNs use pooling layers?

Pooling layers reduce the size of feature maps, making the model faster and less sensitive to small shifts or distortions in the input.

beginner

What role do fully connected layers play in a CNN?

Fully connected layers combine all extracted features to make final predictions, like classifying the image into categories.

intermediate

Explain the concept of 'stride' in convolutional layers.

Stride is how many pixels the filter moves at each step when sliding over the input. Larger strides reduce output size but may skip details.

intermediate

What is the benefit of using multiple convolutional layers stacked together?

Stacking layers lets the CNN learn complex features step-by-step, from simple edges in early layers to detailed shapes in deeper layers.

What does a convolutional filter primarily detect in an image?

A The entire image color

B Local patterns like edges

C Random noise

D Image file size

Convolutional filters scan small parts of the image to detect local patterns such as edges or textures.

Which layer type reduces the spatial size of feature maps in a CNN?

A Dropout layer

B Fully connected layer

C Pooling layer

D Convolutional layer

Pooling layers reduce the size of feature maps to make the model faster and more robust.

What does the 'stride' parameter control in a convolutional layer?

A The number of filters

B The learning rate

C The color channels

D How far the filter moves each step

Stride controls how many pixels the filter moves when sliding over the input.

Why are multiple convolutional layers stacked in a CNN?

A To learn complex features progressively

B To increase image size

C To reduce training data

D To avoid overfitting

Stacking layers helps the model learn from simple to complex features step-by-step.

What is the main function of fully connected layers in CNNs?

A Combine features for final prediction

B Detect edges in images

C Reduce image size

D Normalize input data

Fully connected layers use all extracted features to make the final classification or prediction.

Describe the main components of a CNN architecture and their roles.

Think about how the model processes images step-by-step.

Explain why stacking multiple convolutional layers helps a CNN learn better.

Consider how deeper layers build on earlier layers' outputs.