Benchmark simulation for vgg with depth-wise convolution

vgg_depthwise.py

"""References:

Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for
large-scale image recognition." arXiv preprint arXiv:1409.1556 (2014).
"""
import mxnet as mx

def depthwise_conv(data, kernel, pad, num_filter, name, num_group):
    conv = mx.symbol.Convolution(data=data, kernel=kernel, pad=pad,
        num_filter=num_group, name=name+'_depthwise', num_group=num_group)
    # bn = mx.symbol.BatchNorm(data=conv)
    bn = conv  # for benchmark
    relu = mx.symbol.Activation(data=bn, act_type='relu')
    conv2 = mx.symbol.Convolution(data=relu, kernel=(1, 1), num_filter=num_filter,
        name=name+'_pointwise')
    # bn2 = mx.symbol.BatchNorm(data=conv2)
    bn2 = conv2
    return bn2


def get_symbol(num_classes, **kwargs):
    ## define alexnet
    data = mx.symbol.Variable(name="data")
    # group 1
    conv1_1 = depthwise_conv(data=data, kernel=(3, 3), pad=(1, 1), num_filter=64, name="conv1_1", num_group=3)
    relu1_1 = mx.symbol.Activation(data=conv1_1, act_type="relu", name="relu1_1")
    pool1 = mx.symbol.Pooling(
        data=relu1_1, pool_type="max", kernel=(2, 2), stride=(2,2), name="pool1")
    # group 2
    conv2_1 = depthwise_conv(
        data=pool1, kernel=(3, 3), pad=(1, 1), num_filter=128, name="conv2_1", num_group=64)
    relu2_1 = mx.symbol.Activation(data=conv2_1, act_type="relu", name="relu2_1")
    pool2 = mx.symbol.Pooling(
        data=relu2_1, pool_type="max", kernel=(2, 2), stride=(2,2), name="pool2")
    # group 3
    conv3_1 = depthwise_conv(
        data=pool2, kernel=(3, 3), pad=(1, 1), num_filter=256, name="conv3_1", num_group=128)
    relu3_1 = mx.symbol.Activation(data=conv3_1, act_type="relu", name="relu3_1")
    conv3_2 = depthwise_conv(
        data=relu3_1, kernel=(3, 3), pad=(1, 1), num_filter=256, name="conv3_2", num_group=256)
    relu3_2 = mx.symbol.Activation(data=conv3_2, act_type="relu", name="relu3_2")
    pool3 = mx.symbol.Pooling(
        data=relu3_2, pool_type="max", kernel=(2, 2), stride=(2,2), name="pool3")
    # group 4
    conv4_1 = depthwise_conv(
        data=pool3, kernel=(3, 3), pad=(1, 1), num_filter=512, name="conv4_1", num_group=256)
    relu4_1 = mx.symbol.Activation(data=conv4_1, act_type="relu", name="relu4_1")
    conv4_2 = depthwise_conv(
        data=relu4_1, kernel=(3, 3), pad=(1, 1), num_filter=512, name="conv4_2", num_group=512)
    relu4_2 = mx.symbol.Activation(data=conv4_2, act_type="relu", name="relu4_2")
    pool4 = mx.symbol.Pooling(
        data=relu4_2, pool_type="max", kernel=(2, 2), stride=(2,2), name="pool4")
    # group 5
    conv5_1 = depthwise_conv(
        data=pool4, kernel=(3, 3), pad=(1, 1), num_filter=512, name="conv5_1", num_group=512)
    relu5_1 = mx.symbol.Activation(data=conv5_1, act_type="relu", name="relu5_1")
    conv5_2 = depthwise_conv(
        data=relu5_1, kernel=(3, 3), pad=(1, 1), num_filter=512, name="conv5_2", num_group=512)
    relu5_2 = mx.symbol.Activation(data=conv5_2, act_type="relu", name="conv1_2")
    pool5 = mx.symbol.Pooling(
        data=relu5_2, pool_type="max", kernel=(2, 2), stride=(2,2), name="pool5")
    # group 6
    flatten = mx.symbol.Flatten(data=pool5, name="flatten")
    fc6 = mx.symbol.FullyConnected(data=flatten, num_hidden=4096, name="fc6")
    relu6 = mx.symbol.Activation(data=fc6, act_type="relu", name="relu6")
    drop6 = mx.symbol.Dropout(data=relu6, p=0.5, name="drop6")
    # group 7
    fc7 = mx.symbol.FullyConnected(data=drop6, num_hidden=4096, name="fc7")
    relu7 = mx.symbol.Activation(data=fc7, act_type="relu", name="relu7")
    drop7 = mx.symbol.Dropout(data=relu7, p=0.5, name="drop7")
    # output
    fc8 = mx.symbol.FullyConnected(data=drop7, num_hidden=num_classes, name="fc8")
    softmax = mx.symbol.SoftmaxOutput(data=fc8, name='softmax')
    return softmax