cnn-zalando-paper

cnn-dropout-1 Layer (type) Output Shape Param #

conv2d_1 (Conv2D) (None, 24, 24, 6) 156

max_pooling2d_1 (MaxPooling2 (None, 12, 12, 6) 0

dropout_1 (Dropout) (None, 12, 12, 6) 0

conv2d_2 (Conv2D) (None, 8, 8, 16) 2416

max_pooling2d_2 (MaxPooling2 (None, 4, 4, 16) 0

dropout_2 (Dropout) (None, 4, 4, 16) 0

flatten_1 (Flatten) (None, 256) 0

dense_1 (Dense) (None, 120) 30840

dropout_3 (Dropout) (None, 120) 0

dense_2 (Dense) (None, 84) 10164

dropout_4 (Dropout) (None, 84) 0

dense_3 (Dense) (None, 10) 850

Total params: 44,426
Trainable params: 44,426
Non-trainable params: 0
Train loss/accuracy: 0.21573755418310564/91.87166666666667
Test loss/accuracy: 0.2645787087917328/90.35
6 minutos

cnn-simple-1 Layer (type) Output Shape Param #

conv2d_3 (Conv2D) (None, 24, 24, 6) 156

conv2d_4 (Conv2D) (None, 20, 20, 12) 1812

max_pooling2d_3 (MaxPooling2 (None, 10, 10, 12) 0

dropout_5 (Dropout) (None, 10, 10, 12) 0

flatten_2 (Flatten) (None, 1200) 0

dense_4 (Dense) (None, 90) 108090

dropout_6 (Dropout) (None, 90) 0

dense_5 (Dense) (None, 10) 910

Total params: 110,968
Trainable params: 110,968
Non-trainable params: 0
Train loss/accuracy: 0.04985963567070042/98.91499999999999
Test loss/accuracy: 0.2636391075462103/91.72
7m e 52s

cnn-dropout-2
Layer (type) Output Shape Param #

conv2d_1 (Conv2D) (None, 26, 26, 6) 60

conv2d_2 (Conv2D) (None, 24, 24, 6) 330

max_pooling2d_1 (MaxPooling2 (None, 12, 12, 6) 0

dropout_1 (Dropout) (None, 12, 12, 6) 0

conv2d_3 (Conv2D) (None, 10, 10, 12) 660

conv2d_4 (Conv2D) (None, 8, 8, 12) 1308

max_pooling2d_2 (MaxPooling2 (None, 4, 4, 12) 0

dropout_2 (Dropout) (None, 4, 4, 12) 0

conv2d_5 (Conv2D) (None, 2, 2, 32) 3488

dropout_3 (Dropout) (None, 2, 2, 32) 0

flatten_1 (Flatten) (None, 128) 0

dense_1 (Dense) (None, 120) 15480

dropout_4 (Dropout) (None, 120) 0

dense_2 (Dense) (None, 84) 10164

dropout_5 (Dropout) (None, 84) 0

dense_3 (Dense) (None, 10) 850

Total params: 32,340
Trainable params: 32,340
Non-trainable params: 0
Train loss/accuracy: 0.1973610945492983/92.59666666666668
Test loss/accuracy: 0.2546938347458839/90.81
9 min e 16s

cnn-dropout-3

Layer (type) Output Shape Param #

conv2d_1 (Conv2D) (None, 24, 24, 6) 156

max_pooling2d_1 (MaxPooling2 (None, 12, 12, 6) 0

dropout_1 (Dropout) (None, 12, 12, 6) 0

conv2d_2 (Conv2D) (None, 8, 8, 16) 2416

max_pooling2d_2 (MaxPooling2 (None, 4, 4, 16) 0

dropout_2 (Dropout) (None, 4, 4, 16) 0

flatten_1 (Flatten) (None, 256) 0

dense_1 (Dense) (None, 120) 30840

dropout_3 (Dropout) (None, 120) 0

dense_2 (Dense) (None, 84) 10164

dropout_4 (Dropout) (None, 84) 0

dense_3 (Dense) (None, 10) 850

Total params: 44,426
Trainable params: 44,426
Non-trainable params: 0

Train loss/accuracy: 0.14716978641649087/94.53
Test loss/accuracy: 0.25263977706432345/90.86999999999999
5m e 58s

andreihod / cnn-zalando-paper Goto Github PK