Multiple Convolutions Multiple convolutions Usually there are multiple feature maps, one for each convolution operator
Multiple Convolutions Usually there are multiple feature maps, one for each convolution operator
Non-linearity anh(x) ReLU tanh(a) f(x)= max( 0, x) X+e-x
Non-linearity Tanh(x) ReLU tanh x = e x − e −x e x + e −x 𝑓 𝑥 = max(0, 𝑥)
Pooling Common pooling operations Max pooling reports the maximum output within a rectangular neighborhood Average pooling: reports the average output of a rectangular neighborhood (possibly weighted by the distance from the central pixel) local receptive fields weig sharing pooling input image convolution pooling er
Pooling • Common pooling operations: – Max pooling: reports the maximum output within a rectangular neighborhood. – Average pooling: reports the average output of a rectangular neighborhood (possibly weighted by the distance from the central pixel)
Deep cNN: winner of ImageNet 2012 dense 27128 d 128 Max (Alex et al. 201 2 Multiple feature maps per convolutional layer Multiple convolutional layers for extracting features at different levels Higher-level layers take the feature maps in lower-level layers as input
Deep CNN: winner of ImageNet 2012 • Multiple feature maps per convolutional layer. • Multiple convolutional layers for extracting features at different levels. • Higher-level layers take the feature maps in lower-level layers as input. (Alex et al., 2012)
Deep CNn for Image Classification Classification Click for a Quick Example Maximally accurate Maximally specific cat feline 1.74269 domestic cat 1.70760 tabby 0.94807 domestic animal CNN took 0.064 seconds Try out a live demo at http://demo.caffeberkeleyvision.org
Deep CNN for Image Classification Try out a live demo at http://demo.caffe.berkeleyvision.org/