Pixel

[parthraghav]

[cazala]

1. pixel function return an array with 3 values, (the red, green and blue components of a pixel). ox and oy are offsets. There is an index variable that iterates over all the pixels, the offsets let you get the values of the pixels next to the index pixel.
2. px is the input of the network, is an array with the rgb values of 9 pixels. It's like if the network could 'see' only 9 pixels of the image at a time, instead of feeding the whole image each activation. This let's you use a way smaller network, only 9 * 3 (pixels * rgb values) = 27 inputs. Otherwise for a 125px * 125px you would need an network with 125 * 125 * 3 = 46875 inputs. You could also make the network 'see' only 1 pixel at a time, that's enough for certain filters like greyscale, but that cannot solve some of the filters that need more context information, like blur.