Conv2D layers in TensorFlow - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is a Conv2D layer in neural networks?

A Conv2D layer is a type of layer that applies 2D convolution filters to input images or feature maps to detect patterns like edges or shapes.

beginner

What do the 'filters' and 'kernel_size' parameters control in a Conv2D layer?

'filters' sets how many different patterns the layer will learn, and 'kernel_size' sets the size of the small window that scans over the input.

intermediate

How does padding affect the output size of a Conv2D layer?

Padding adds extra pixels around the input edges. 'same' padding keeps output size the same as input, while 'valid' padding means no padding and smaller output.

intermediate

What role does the stride parameter play in a Conv2D layer?

Stride controls how many pixels the filter moves each step. Larger strides reduce output size by skipping positions.

beginner

Why are Conv2D layers important for image recognition tasks?

Conv2D layers automatically learn important features like edges and textures, making them very good at understanding images without manual feature design.

What does the 'filters' parameter in Conv2D specify?

A Stride length

B Size of the input image

C Number of feature detectors

D Number of output classes

'filters' sets how many different patterns the Conv2D layer will learn to detect.

What does 'same' padding do in a Conv2D layer?

A Removes border pixels

B Keeps output size same as input

C Doubles the output size

D No padding added

'same' padding adds pixels so output height and width match input.

If stride is set to 2, what happens to the output size?

A Output size halves

B Output size doubles

C Output size stays same

D Output size becomes zero

Stride 2 moves filter 2 pixels at a time, reducing output size roughly by half.

What is the kernel_size in Conv2D?

A Batch size

B Number of filters

C Number of input channels

D Size of the filter window

Kernel size is the height and width of the filter window scanning the input.

Which layer type is best for detecting edges in images?

A Conv2D

B Dense

C Dropout

D Flatten

Conv2D layers learn filters that detect edges and patterns in images.

Explain how a Conv2D layer processes an input image step-by-step.

Think about how the filter moves and what it produces.

Describe the effect of changing the number of filters and kernel size in a Conv2D layer.

Consider what each parameter controls in pattern detection.