發布於 2018-04-15
觀看次數: 39
  • 00:21 1.
    Convolutional Neural Network
  • 02:54 2.
    Why CNN for Image
  • 01:59 3.
    Why CNN for Image
  • 01:05 4.
    Why CNN for Image
  • 05:28 5.
    The whole CNN
  • 01:15 6.
    The whole CNN
  • 00:03 7.
    The whole CNN
  • 01:58 8.
    CNN – Convolution
  • 02:08 9.
    CNN – Convolution
  • 00:57 10.
    CNN – Convolution
  • 02:00 11.
    CNN – Convolution
  • 00:33 12.
    CNN – Convolution
  • 02:14 13.
    CNN – Colorful image
  • 01:07 14.
    Convolution v.s. Fully Connected
  • 02:41 15.
    3
  • 01:23 16.
    3
  • 00:03 17.
    The whole CNN
  • 01:06 18.
    CNN – Max Pooling
  • 01:48 19.
    CNN – Max Pooling
  • 00:38 20.
    The whole CNN
  • 00:07 21.
    The whole CNN
  • 00:11 22.
    Flatten
  • 01:57 23.
    input
  • 00:00 24.
    input
  • 00:02 25.
    input
  • 07:20 26.
    input
  • 01:25 27.
    input
  • 03:33 28.
    What does machine learn?
  • 01:57 29.
    First Convolution Layer
  • 03:19 30.
    How about higher layers?
  • 02:38 31.
  • 00:32 32.
  • 03:01 33.
    Find an image maximizing the output of neuron:
  • 00:02 34.
    Can we see digits?
  • 00:15 35.
    Find an image maximizing the output of neuron:
  • 03:45 36.
  • 00:35 37.
    Find an image maximizing the output of neuron:
  • 00:00 38.
    Can we see digits?
  • 00:02 39.
    Find an image maximizing the output of neuron:
  • 00:00 40.
    Can we see digits?
  • 00:11 41.
    Find an image maximizing the output of neuron:
  • 00:15 42.
    Can we see digits?
  • 00:04 43.
    Find an image maximizing the output of neuron:
  • 01:09 44.
    Can we see digits?
  • 01:38 45.
    What does CNN learn?
  • 02:27 46.
    Slide 35
  • 02:52 47.
  • 01:05 48.
    Slide 37
  • 00:56 49.
    Deep Dream
  • Index
  • Note
  • Discuss
  • Fullscreen
W5a_CNN
1:13:26, 發布於 2018-04-15 by 許光習