votes=1

[data]
type=data
outputs=data
channels=1
height=21
width=101

[rotate1]
type=rotate
inputs=data
outputs=rotate1

##### layer 1 #####

[conv1]
type=conv
inputs=rotate1
outputs=conv1
filters=12
windowSize=3
windowStride=1
padding=0
weights=conv1_0.dat
biases=conv1_1.dat

[bn1]
type=batch_norm
inputs=conv1
outputs=conv1
means=bn1_0.dat
variances=bn1_1.dat
nums=bn1_2.dat

[conv1-relu]
type=neuron
inputs=conv1
outputs=conv1
neuronType=RELU

###### layer 2 ######

[conv2]
type=conv
inputs=conv1
outputs=conv2
filters=12
windowSize=3
windowStride=1
padding=0
weights=conv2_0.dat
biases=conv2_1.dat

[bn2]
type=batch_norm
inputs=conv2
outputs=conv2
means=bn2_0.dat
variances=bn2_1.dat
nums=bn2_2.dat

[conv2-relu]
type=neuron
inputs=conv2
outputs=conv2
neuronType=RELU

[pool2]
type=pool
inputs=conv2
outputs=pool2
windowSize=2
windowStride=2
padding=0
poolType=MAX

###### layer 3 ######

[conv3]
type=conv
inputs=pool2
outputs=conv3
filters=16
windowSize=3
windowStride=1
padding=0
weights=conv3_0.dat
biases=conv3_1.dat

[bn3]
type=batch_norm
inputs=conv3
outputs=conv3
means=bn3_0.dat
variances=bn3_1.dat
nums=bn3_2.dat

[conv3-relu]
type=neuron
inputs=conv3
outputs=conv3
neuronType=RELU

###### layer 4 ######

[conv4]
type=conv
inputs=conv3
outputs=conv4
filters=16
windowSize=3
windowStride=1
padding=0
weights=conv4_0.dat
biases=conv4_1.dat

[bn4]
type=batch_norm
inputs=conv4
outputs=conv4
means=bn4_0.dat
variances=bn4_1.dat
nums=bn4_2.dat

[conv4-relu]
type=neuron
inputs=conv4
outputs=conv4
neuronType=RELU

###### layer 5 ######

[conv5]
type=conv
inputs=conv4
outputs=conv5
filters=24
windowSize=3
windowStride=1
padding=0
weights=conv5_0.dat
biases=conv5_1.dat

[bn5]
type=batch_norm
inputs=conv5
outputs=conv5
means=bn5_0.dat
variances=bn5_1.dat
nums=bn5_2.dat

[conv5-relu]
type=neuron
inputs=conv5
outputs=conv5
neuronType=RELU

###### layer 6 ######

[conv6]
type=conv
inputs=conv5
outputs=conv6
filters=24
windowSize=3
windowStride=1
padding=0
weights=conv6_0.dat
biases=conv6_1.dat

[bn6]
type=batch_norm
inputs=conv6
outputs=conv6
means=bn6_0.dat
variances=bn6_1.dat
nums=bn6_2.dat

[conv6-relu]
type=neuron
inputs=conv6
outputs=conv6
neuronType=RELU

###### Final layer ######

[rotate6]
type=rotate
inputs=conv6
outputs=rotate6

[fc]
type=fc
inputs=rotate6
outputs=fc
filters=25
weights=fc_0.dat
biases=fc_1.dat

[prob]
type=softmax
inputs=fc
outputs=prob