en
0.25
0.5
0.75
1.25
1.5
1.75
2
Visualizing and Understanding Convolutional Networks
Published on Oct 29, 201416481 Views
Large Convolutional Network models have recently demonstrated impressive classification performance on the ImageNet benchmark Krizhevsky et al. [18]. However there is no clear understanding of why the
Related categories
Chapter list
Visualization and Understanding Convolutional Neural Networks00:00
Overview - 100:14
Convnets Show Huge Gains00:33
Convolutional Networks (LeCun et al. ’89)01:01
Overview - 202:01
Deconvolutional Networks02:12
Reversible Max Pooling02:59
Layer 1 Filters04:00
Projecting back from Higher Layers04:22
Visualizations of Higher Layers05:13
Layer 1: Top-9 Patches05:46
Layer 2: Top-106:21
Layer 2: Top-907:39
Layer 2: Top-9 Patches07:40
Layer 3: Top-107:41
Layer 3: Top-908:35
Layer 3: Top-9 Patches08:39
Layer 4: Top-108:45
Layer 4: Top-909:03
Layer 4: Top-9 Patches09:25
Layer 5: Top-109:35
Layer 5: Top‐910:26
Layer 5: Top‐9 Patches10:31
Occlusion Experiment - 110:32
Occlusion Experiment - 211:03
Occlusion Experiment - 311:41
Occlusion Experiment - 412:17
Occlusion Experiment - 512:32
Occlusion Experiment - 613:01
Occlusion Experiment - 713:32
Lack of Understanding13:39
Visualizations Help – 2% Boost13:40
ImageNet ClassificaMon 2013 Results14:31
Recent Success14:43
Overview - 315:01
Caltech 256 - 115:04
Caltech 256 - 215:09
Summary15:18
Thanks!15:30