Open junxnone opened 5 years ago
Normal Cell
Reduction Cell
RL
Policy Gradient
Model
Normal/Reduction Cell
Block
在 paper 中 B = 5
hidden states
Operation
B Block
1x7 then 7x1 convolution 3x3 dilated convolution 3x3 average pooling 3x3 max pooling 5x5 max pooling 7x7 max pooling 1x1 convolution 3x3 convolution 3x3 depthwise-separable conv 5x5 depthwise-seperable conv 7x7 depthwise-separable conv
NASNet-A/B/C
Layer Normalization
Instance Normalization
搜素到的相对比较好的 Normal Cell & Reduction Cell - NASNet-A -- |
Layer Normalization Instance Normalization
junxnone/tech-io#107 junxnone/tech-io#805
NASNet
Normal Cell
- 输入和输出的特征图维度相同Reduction Cell
- 输出特征图长宽减半RL
+Policy Gradient
Arch
Model
=Normal Cell
x N +Reduction Cell
x MNormal/Reduction Cell
=Block
x BBlock
=hidden states
x 2 +Operation
x 3B Block
/ Each Block predict 5 Steps1x7 then 7x1 convolution 3x3 dilated convolution 3x3 average pooling 3x3 max pooling 5x5 max pooling 7x7 max pooling 1x1 convolution 3x3 convolution 3x3 depthwise-separable conv 5x5 depthwise-seperable conv 7x7 depthwise-separable conv
Search Result
NASNet-A/B/C
- 最后没有Concatenate
-
Layer Normalization
&Instance Normalization
-
Layer Normalization
&Instance Normalization
搜素到的相对比较好的
Normal Cell
&Reduction Cell
- NASNet-A -- |Reference