def increase_b(x, scale, name=None):
pad = 'same'
branch0 = conv2d.build(x, 192, 1, 1, pad, True, name=name + 'b0')
branch1 = conv2d.build(x, 128, 1, 1, pad, True, name=name + 'b1_1')
branch1 = conv2d.build(branch1,
160, [1, 7],
1,
pad,
True,
name=name + 'b1_2')
branch1 = conv2d.build(branch1,
192, [7, 1],
1,
pad,
True,
name=name + 'b1_3')
branches = [branch0, branch1]
mixed = Concatenate(axis=3, name=name + '_mixed')(branches)
filt_exp_1x1 = conv2d.build(mixed,
1152,
1,
1,
pad,
False,
name=name + 'filt_exp_1x1')
final_lay = Lambda(scaling,
output_shape=backend.int_shape(x)[1:],
arguments={'scale': scale},
name=name + 'act_scaling')([x, filt_exp_1x1])
return final_lay
def reduction_a(x):
# 35 × 35 to 17 × 17 reduction module.
x_red_11 = MaxPooling2D(3, strides=2, padding='valid', name='red_maxpool_1')(x)
x_red_12 = conv2d.build(x, 384, 3, 2, 'valid', True, name='x_red1_c1')
x_red_13 = conv2d.build(x, 256, 1, 1, 'same', True, name='x_red1_c2_1')
x_red_13 = conv2d.build(x_red_13, 256, 3, 1, 'same', True, name='x_red1_c2_2')
x_red_13 = conv2d.build(x_red_13, 384, 3, 2, 'valid', True, name='x_red1_c2_3')
x = Concatenate(axis=3, name='red_concat_1')([x_red_11, x_red_12, x_red_13])
return x
def increase_a(x,scale,name=None):
pad = 'same'
branch0 = conv2d.build(x,32,1,1,pad,True,name=name+'b0')
branch1 = conv2d.build(x,32,1,1,pad,True,name=name+'b1_1')
branch1 = conv2d.build(branch1,32,3,1,pad,True,name=name+'b1_2')
branch2 = conv2d.build(x,32,1,1,pad,True,name=name+'b2_1')
branch2 = conv2d.build(branch2,48,3,1,pad,True,name=name+'b2_2')
branch2 = conv2d.build(branch2,64,3,1,pad,True,name=name+'b2_3')
branches = [branch0,branch1,branch2]
mixed = Concatenate(axis=3, name=name + '_concat')(branches)
filt_exp_1x1 = conv2d.build(mixed,384,1,1,pad,False,name=name+'filt_exp_1x1')
final_lay = Lambda(scaling,
output_shape=backend.int_shape(x)[1:],
arguments={'scale': scale},
name=name+'act_scaling')([x, filt_exp_1x1])
return final_lay
def reduction_b(x):
# 17 × 17 to 8 × 8 reduction module.
x_red_21 = MaxPooling2D(3,
strides=2,
padding='valid',
name='red_maxpool_2')(x)
x_red_22 = conv2d.build(x, 256, 1, 1, 'same', True, name='x_red2_c11')
x_red_22 = conv2d.build(x_red_22,
384,
3,
2,
'valid',
True,
name='x_red2_c12')
x_red_23 = conv2d.build(x, 256, 1, 1, 'same', True, name='x_red2_c21')
x_red_23 = conv2d.build(x_red_23,
256,
3,
2,
'valid',
True,
name='x_red2_c22')
x_red_24 = conv2d.build(x, 256, 1, 1, 'same', True, name='x_red2_c31')
x_red_24 = conv2d.build(x_red_24,
256,
3,
1,
'same',
True,
name='x_red2_c32')
x_red_24 = conv2d.build(x_red_24,
256,
3,
2,
'valid',
True,
name='x_red2_c33')
x = Concatenate(
axis=3, name='red_concat_2')([x_red_21, x_red_22, x_red_23, x_red_24])
return x
def build(input):
x = conv2d.build(input, 32, 3, 2, 'valid', True, name='conv1')
x = conv2d.build(x, 32, 3, 1, 'valid', True, name='conv2')
x = conv2d.build(x, 64, 3, 1, 'valid', True, name='conv3')
x_11 = MaxPooling2D(3,
strides=1,
padding='valid',
name='stem_br_11' + '_maxpool_1')(x)
x_12 = conv2d.build(x, 64, 3, 1, 'valid', True, name='stem_br_12')
x = Concatenate(axis=3, name='stem_concat_1')([x_11, x_12])
x_21 = conv2d.build(x, 64, 1, 1, 'same', True, name='stem_br_211')
x_21 = conv2d.build(x_21, 64, [1, 7], 1, 'same', True, name='stem_br_212')
x_21 = conv2d.build(x_21, 64, [7, 1], 1, 'same', True, name='stem_br_213')
x_21 = conv2d.build(x_21, 96, 3, 1, 'valid', True, name='stem_br_214')
x_22 = conv2d.build(x, 64, 1, 1, 'same', True, name='stem_br_221')
x_22 = conv2d.build(x_22, 96, 3, 1, 'valid', True, name='stem_br_222')
x = Concatenate(axis=3, name='stem_concat_2')([x_21, x_22])
x_31 = conv2d.build(x, 192, 3, 1, 'valid', True, name='stem_br_31')
x_32 = MaxPooling2D(3,
strides=1,
padding='valid',
name='stem_br_32' + '_maxpool_2')(x)
x = Concatenate(axis=3, name='stem_concat_3')([x_31, x_32])
return x