| """ |
| |
| Inception + BN, suitable for images with around 224 x 224 |
| |
| Reference: |
| |
| Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep |
| network training by reducing internal covariate shift. arXiv preprint |
| arXiv:1502.03167, 2015. |
| |
| """ |
| import mxnet as mx |
| |
| eps = 1e-10 + 1e-5 |
| bn_mom = 0.9 |
| fix_gamma = False |
| |
| |
| def ConvFactory(data, num_filter, kernel, stride=(1,1), pad=(0, 0), name=None, suffix='', attr={}): |
| conv = mx.symbol.Convolution(data=data, num_filter=num_filter, kernel=kernel, stride=stride, pad=pad, name='conv_%s%s' %(name, suffix)) |
| bn = mx.symbol.BatchNorm(data=conv, fix_gamma=fix_gamma, eps=eps, momentum=bn_mom, name='bn_%s%s' %(name, suffix)) |
| act = mx.symbol.Activation(data=bn, act_type='relu', name='relu_%s%s' %(name, suffix), attr=attr) |
| return act |
| |
| def InceptionFactoryA(data, num_1x1, num_3x3red, num_3x3, num_d3x3red, num_d3x3, pool, proj, name): |
| # 1x1 |
| c1x1 = ConvFactory(data=data, num_filter=num_1x1, kernel=(1, 1), name=('%s_1x1' % name)) |
| # 3x3 reduce + 3x3 |
| c3x3r = ConvFactory(data=data, num_filter=num_3x3red, kernel=(1, 1), name=('%s_3x3' % name), suffix='_reduce') |
| c3x3 = ConvFactory(data=c3x3r, num_filter=num_3x3, kernel=(3, 3), pad=(1, 1), name=('%s_3x3' % name)) |
| # double 3x3 reduce + double 3x3 |
| cd3x3r = ConvFactory(data=data, num_filter=num_d3x3red, kernel=(1, 1), name=('%s_double_3x3' % name), suffix='_reduce') |
| cd3x3 = ConvFactory(data=cd3x3r, num_filter=num_d3x3, kernel=(3, 3), pad=(1, 1), name=('%s_double_3x3_0' % name)) |
| cd3x3 = ConvFactory(data=cd3x3, num_filter=num_d3x3, kernel=(3, 3), pad=(1, 1), name=('%s_double_3x3_1' % name)) |
| # pool + proj |
| pooling = mx.symbol.Pooling(data=data, kernel=(3, 3), stride=(1, 1), pad=(1, 1), pool_type=pool, name=('%s_pool_%s_pool' % (pool, name))) |
| cproj = ConvFactory(data=pooling, num_filter=proj, kernel=(1, 1), name=('%s_proj' % name)) |
| # concat |
| concat = mx.symbol.Concat(*[c1x1, c3x3, cd3x3, cproj], name='ch_concat_%s_chconcat' % name) |
| return concat |
| |
| def InceptionFactoryB(data, num_3x3red, num_3x3, num_d3x3red, num_d3x3, name): |
| # 3x3 reduce + 3x3 |
| c3x3r = ConvFactory(data=data, num_filter=num_3x3red, kernel=(1, 1), name=('%s_3x3' % name), suffix='_reduce') |
| c3x3 = ConvFactory(data=c3x3r, num_filter=num_3x3, kernel=(3, 3), pad=(1, 1), stride=(2, 2), name=('%s_3x3' % name)) |
| # double 3x3 reduce + double 3x3 |
| cd3x3r = ConvFactory(data=data, num_filter=num_d3x3red, kernel=(1, 1), name=('%s_double_3x3' % name), suffix='_reduce') |
| cd3x3 = ConvFactory(data=cd3x3r, num_filter=num_d3x3, kernel=(3, 3), pad=(1, 1), stride=(1, 1), name=('%s_double_3x3_0' % name)) |
| cd3x3 = ConvFactory(data=cd3x3, num_filter=num_d3x3, kernel=(3, 3), pad=(1, 1), stride=(2, 2), name=('%s_double_3x3_1' % name)) |
| # pool + proj |
| pooling = mx.symbol.Pooling(data=data, kernel=(3, 3), stride=(2, 2), pad=(1, 1), pool_type="max", name=('max_pool_%s_pool' % name)) |
| # concat |
| concat = mx.symbol.Concat(*[c3x3, cd3x3, pooling], name='ch_concat_%s_chconcat' % name) |
| return concat |
| |
| # A Simple Downsampling Factory |
| def DownsampleFactory(data, ch_3x3, name, attr): |
| # conv 3x3 |
| conv = ConvFactory(data=data, name=name+'_conv',kernel=(3, 3), stride=(2, 2), num_filter=ch_3x3, pad=(1, 1), attr=attr) |
| # pool |
| pool = mx.symbol.Pooling(data=data, name=name+'_pool',kernel=(3, 3), stride=(2, 2), pad=(1, 1), pool_type='max', attr=attr) |
| # concat |
| concat = mx.symbol.Concat(*[conv, pool], name=name+'_ch_concat') |
| return concat |
| |
| # A Simple module |
| def SimpleFactory(data, ch_1x1, ch_3x3, name, attr): |
| # 1x1 |
| conv1x1 = ConvFactory(data=data, name=name+'_1x1', kernel=(1, 1), pad=(0, 0), num_filter=ch_1x1, attr=attr) |
| # 3x3 |
| conv3x3 = ConvFactory(data=data, name=name+'_3x3', kernel=(3, 3), pad=(1, 1), num_filter=ch_3x3, attr=attr) |
| #concat |
| concat = mx.symbol.Concat(*[conv1x1, conv3x3], name=name+'_ch_concat') |
| return concat |
| |
| |
| def get_symbol(num_classes, image_shape, **kwargs): |
| image_shape = [int(l) for l in image_shape.split(',')] |
| (nchannel, height, width) = image_shape |
| # attr = {'force_mirroring': 'true'} |
| attr = {} |
| |
| # data |
| data = mx.symbol.Variable(name="data") |
| if height <= 28: |
| # a simper version |
| conv1 = ConvFactory(data=data, kernel=(3,3), pad=(1,1), name="1", num_filter=96, attr=attr) |
| in3a = SimpleFactory(conv1, 32, 32, 'in3a', attr) |
| in3b = SimpleFactory(in3a, 32, 48, 'in3b', attr) |
| in3c = DownsampleFactory(in3b, 80, 'in3c', attr) |
| in4a = SimpleFactory(in3c, 112, 48, 'in4a', attr) |
| in4b = SimpleFactory(in4a, 96, 64, 'in4b', attr) |
| in4c = SimpleFactory(in4b, 80, 80, 'in4c', attr) |
| in4d = SimpleFactory(in4c, 48, 96, 'in4d', attr) |
| in4e = DownsampleFactory(in4d, 96, 'in4e', attr) |
| in5a = SimpleFactory(in4e, 176, 160, 'in5a', attr) |
| in5b = SimpleFactory(in5a, 176, 160, 'in5b', attr) |
| pool = mx.symbol.Pooling(data=in5b, pool_type="avg", kernel=(7,7), name="global_pool", attr=attr) |
| else: |
| # stage 1 |
| conv1 = ConvFactory(data=data, num_filter=64, kernel=(7, 7), stride=(2, 2), pad=(3, 3), name='1') |
| pool1 = mx.symbol.Pooling(data=conv1, kernel=(3, 3), stride=(2, 2), name='pool_1', pool_type='max') |
| # stage 2 |
| conv2red = ConvFactory(data=pool1, num_filter=64, kernel=(1, 1), stride=(1, 1), name='2_red') |
| conv2 = ConvFactory(data=conv2red, num_filter=192, kernel=(3, 3), stride=(1, 1), pad=(1, 1), name='2') |
| pool2 = mx.symbol.Pooling(data=conv2, kernel=(3, 3), stride=(2, 2), name='pool_2', pool_type='max') |
| # stage 2 |
| in3a = InceptionFactoryA(pool2, 64, 64, 64, 64, 96, "avg", 32, '3a') |
| in3b = InceptionFactoryA(in3a, 64, 64, 96, 64, 96, "avg", 64, '3b') |
| in3c = InceptionFactoryB(in3b, 128, 160, 64, 96, '3c') |
| # stage 3 |
| in4a = InceptionFactoryA(in3c, 224, 64, 96, 96, 128, "avg", 128, '4a') |
| in4b = InceptionFactoryA(in4a, 192, 96, 128, 96, 128, "avg", 128, '4b') |
| in4c = InceptionFactoryA(in4b, 160, 128, 160, 128, 160, "avg", 128, '4c') |
| in4d = InceptionFactoryA(in4c, 96, 128, 192, 160, 192, "avg", 128, '4d') |
| in4e = InceptionFactoryB(in4d, 128, 192, 192, 256, '4e') |
| # stage 4 |
| in5a = InceptionFactoryA(in4e, 352, 192, 320, 160, 224, "avg", 128, '5a') |
| in5b = InceptionFactoryA(in5a, 352, 192, 320, 192, 224, "max", 128, '5b') |
| # global avg pooling |
| pool = mx.symbol.Pooling(data=in5b, kernel=(7, 7), stride=(1, 1), name="global_pool", pool_type='avg') |
| |
| # linear classifier |
| flatten = mx.symbol.Flatten(data=pool) |
| fc1 = mx.symbol.FullyConnected(data=flatten, num_hidden=num_classes) |
| softmax = mx.symbol.SoftmaxOutput(data=fc1, name='softmax') |
| return softmax |
