Couldnt understand receptive field

#1

can u please tell me the exact meaning of receptive field??

#2

Hi the receptive field means the area or pixels of input image which are covered under another layes. For eg suppose there are 2 conv layers 1&2 and one input image. 1st conv layer will look at the 3*3=9 pixels of imput image as the kernel used is of shape (3,3) and we used (3,3) portion of input image and element wise multiplied it by kernel (3,3) and stored the result in one pixel by adding them in activation map. So effectively one pixel in 1st layer of convolution or activation map is looking at (3,3) arial pixels of input image. Now for 2nd conv layer, it is looking are (3,3) arial pixels of 1st conv layer. But in the 1st layer the second adjacent pixel is not looking to new (3,3) area of input image as there is stride of 1. During sliding of kernel over the input image some area of pixels is common on the input image. So for second pixel in 1st conv layer, (3,2) arial pixels are common of input image. Only 3 new pixels are being seen (3,1).
Note: (3,2) + (3,1) =(3,3)
If the stride would have been of 3, then each pixel in actiovation maps would be looking at new set of 3,3 pixels of input image.

#3

correction :**Now for second pixel in first conv layer

#4

Therefore receptive field is like the new non-common pixels which are seen by one pixel in later layer.

#5

I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.

On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.