No content

LONG ET AL., INTRO • Fully convolutional networks 

LONG ET AL., INTRO 

LONG ET AL., INTRO • Q: How do you make a network fully convolutional? • A: by making it fully convolutional 

LONG ET AL., INTRO • Okay, but how do we obtain “dense” predictions, i.e., predictions for every pixel in the output? 1. Shift and stitch, or equivalently ‘a trous’ / dilated convolution 2. Upsampling, AKA backwards convolution or deconvolution 

LONG ET AL., INTRO 

LONG ET AL., RESULTS • Converting classification nets to segmentation nets yielded state-of-the-art results 

LONG ET AL., RESULTS • Adding the “deep jet” with skip layers improved the segmentation detail 

VAN DEN OORD ET AL.: INTRO • A generative model for raw audio – “What if we used PixelCNN on audio data?” 

VAN DEN OORD ET AL.: INTRO • Even more secret ingredient: dilated causal convolution 

VAN DEN OORD ET AL.: INTRO • Even more secret ingredient: dilated causal convolution 

VAN DEN OORD ET AL.: INTRO • Yet more secret ingredients: – Output is a softmax layer trained on transformed data • non-linear transformation that can be mapped back to full range of 16-bit audio output – Gated activation units – Residual and skip connections 

VAN DEN OORD ET AL.: INTRO • Your model needs a conditioner – global conditioning – local conditioning 

VAN DEN OORD ET AL.: RESULTS • 3.1: We got it to make up speech. • 3.2: It did better than other models on text to speech (TTS) – Other models are concatenative (LSTM-RNN) and parameterized (HMM) 

VAN DEN OORD ET AL.: RESULTS • 3.2: It did better than other models on text to speech (TTS) (cont.) • 3.3: We got it to make music.