conv neural network - what should be the depth of a convolution layer for grayscale and black and white images? -
i'm going through lecture notes here:
http://cs231n.github.io/convolutional-networks/
in first convolution layer, typically @ 5x5x3 3 refers rgb color space , 5x5 height , width of picture.
however, if i'm looking @ grayscale images 5x5x1 last dimension 0 1 (perfectly black white)? similarly, if simpler pure black , white images, 5x5x1 last dimension 0 or 1?
yes you're right. in case of grayscale or black , white images have 1 feature map in input layer.
Comments
Post a Comment