Search code examples
pythonconv-neural-network

Convolution Neural Network and Receptive field

could someone please explain how to do the below question

Suppose a 3x3 convolution kernel is used in the first layer, a 2x2 convolution kernel is used in the second layer and a 4x4 convolution kernel is used in the third layer. What is the receptive field of the element of the output of the third layer?

is there a general equation to solve the above problem? im relatively new to CNN so any help is appreciated

thanks in advance!

i did read the dive into deep learining book but could understand anything

Solution

  • From: https://theaisummer.com/receptive-field/

    For two sequential convolutional layers 2,1 f2,f1 with kernel size k, stride s, receptive field r:

    r1=s2×r2+(k2−s2)

    Although, I don't know how useful that really is for you... Think of receptive field like looking through a straw with your own eyes... A really narrow straw will limit your "receptive field" making it more difficult to actually discern what you are looking at or seeing